Anthropic vs OpenAI: The AI Bubble Race, and Who Survives If It Pops

Two companies sit at the center of every argument about whether artificial intelligence is a bubble. OpenAI, the maker of ChatGPT, has the bigger consumer brand, the larger user base and the most audacious infrastructure plans in corporate history. Anthropic, the maker of Claude, was founded in 2021 by people who left OpenAI. It has grown faster than almost anyone predicted, mostly by selling to businesses and software developers, and as of late October 2026 it is the one with an IPO prospectus on file.

Between them, the two labs have raised well over $250 billion in private capital since 2023, been valued at close to $2 trillion combined, and signed compute contracts that run past a trillion dollars. Both still lose money on a GAAP basis. Both say they will be enormously profitable later. If the AI boom keeps going, those bets may look visionary. If it deflates, with spending slowing, prices for AI services collapsing or investors losing patience, the question becomes very concrete: which of these companies can pay its bills?

This article compares the two head to head, using dated, reported figures. Sibling articles on Apiary cover the general AI bubble debate and overall chatbot market share. Here the focus is narrower: money in, money out, compute owed, who is in charge, and how each company would fare in a downturn. Because both labs buy most of their chips, directly or through cloud partners, from one company, there is also a long section on Nvidia and what an AI slowdown would mean for gamers who buy GeForce RTX cards.

Disclaimer: This article is for general information only and is not financial or investment advice.

A note on numbers: neither company was public for most of this period, so many figures come from leaked investor documents reported by outlets such as The Information, the Financial Times, Bloomberg, Reuters and the Wall Street Journal. Where a figure is a projection or a third-party estimate, it is labeled as such. "Run-rate" (annualized) revenue is not the same as revenue actually booked over a year. The Financial Times and Bloomberg have both noted that OpenAI and Anthropic calculate run-rate differently, so direct comparisons should be read with caution.

The short version

• Valuation: OpenAI closed a $122 billion round at an $852 billion post-money valuation on March 31, 2026. Anthropic raised $65 billion at a $965 billion post-money valuation in its Series H, announced around May 28 to June 1, 2026. On paper, that made Anthropic the more valuable private company. Press reports in September 2026 said OpenAI had discussed terms that could value it near $1.2 trillion, and coverage of Anthropic's IPO has floated valuations around $2 trillion. Neither of those figures had been confirmed by a priced deal as of October 31, 2026.
• Revenue: Anthropic told investors its run-rate passed $65 billion at the end of July 2026. OpenAI's run-rate topped $40 billion in August 2026, according to Bloomberg, and was about $50 billion in September, according to the Financial Times. On reported half-year revenue, Anthropic's S-1 shows $16.3 billion in the first half of 2026. OpenAI reported $5.7 billion in the first quarter alone, according to The Information.
• Burn: OpenAI's own projections, as reported by the FT in September 2026, show about $278 billion of negative free cash flow between 2026 and 2030 and current cash running out by 2028 without new funding. Anthropic's S-1 shows its first quarter of operating profit ($974 million in Q2 2026), but a huge first-half net loss of $80.7 billion, most of it a non-cash accounting charge.
• Compute: OpenAI's announced infrastructure deals, including Stargate, Oracle, Nvidia, AMD, Broadcom, CoreWeave, AWS and Azure, are widely tallied at $1 trillion to $1.4 trillion or more over roughly a decade. Anthropic's named contracts in its S-1 total roughly $637 billion, by one analyst's sum.
• Structure: OpenAI became a public benefit corporation (OpenAI Group PBC) controlled by a nonprofit foundation in October 2025, with Microsoft holding about 27%. Anthropic has been a public benefit corporation since its early days and gives a Long-Term Benefit Trust the power to elect a majority of its board.
• Bubble odds: Anthropic's recent numbers point to a shorter path to self-funding. OpenAI has a far larger consumer audience, about 1 billion weekly ChatGPT users as of August 2026, and more ways to make money from it, including ads. Both are priced for spectacular growth, and both have long-term compute bills that only make sense if demand keeps compounding.

Funding rounds and valuations, 2023 to October 2026

Big Tech money underpins both companies, and the money usually comes paired with compute. Microsoft backed OpenAI for years. Amazon and Google became Anthropic's anchors. By 2026, though, the lines had blurred: Amazon and Nvidia invested in OpenAI's 2026 round, while Microsoft and Nvidia invested in Anthropic.

OpenAI: major rounds

Anthropic: major rounds

Valuation over time (post-money, $ billions)

Each block is roughly $25 billion.

OpenAI
Oct 2024  $157B  ██████
Mar 2025  $300B  ████████████
Oct 2025  $500B  ████████████████████
Mar 2026  $852B  ██████████████████████████████████

Anthropic
Mar 2025  $61.5B ██
Sep 2025  $183B  ███████
Feb 2026  $380B  ███████████████
May 2026  $965B  ██████████████████████████████████████

Look at the slope. Anthropic's valuation rose roughly fifteenfold in about 14 months, from $61.5 billion in March 2025 to $965 billion in May 2026. OpenAI's rose about 5.4 times over a slightly longer stretch, from $157 billion in October 2024 to $852 billion in March 2026. Bubble skeptics point to exactly this kind of curve. Bulls answer that revenue grew even faster, as the next section shows.

Revenue and annualized run-rates

OpenAI's revenue path

In January 2026, OpenAI said its annualized revenue had grown from about $2 billion in 2023 to $6 billion in 2024 and more than $20 billion in 2025. Its compute capacity grew from 0.2 gigawatts to 0.6 and then about 1.9 gigawatts over the same three years. Actual recognized revenue for 2025 was about $13.1 billion, according to internal documents reported by The Information. The run-rate figure is higher because revenue was still climbing steeply at year-end.

In 2026:

• Q1 2026: $5.7 billion in revenue and $3.7 billion in cash burn, both roughly triple a year earlier. OpenAI ended the quarter with more than $73 billion in cash and marketable securities, up from about $40 billion at the end of 2025 (The Information, via PYMNTS, June 2026).
• August 2026: Run-rate above $40 billion, roughly double the late-2025 pace. An internal note from president Greg Brockman said the monthly run-rate rose more than 20% in July alone (Bloomberg, August 13, 2026).
• September 2026: Some outlets estimated OpenAI was near a $70 billion run-rate. The Financial Times reported on October 8, 2026, that the actual figure was about $50 billion as of September. It said the $70 billion number came from investors applying a 70% growth figure to the August estimate.
• Full-year 2026 projection: About $36 billion in revenue, according to investor materials reported by the FT in September 2026. That rises to a projected $350 billion in 2030.

Anthropic's revenue path

Anthropic's growth has been steeper. In mid-2025, CEO Dario Amodei said the company "went from zero to $100 million in 2023" on a run-rate basis, reached about $1 billion in 2024, and was well above $4 billion by mid-2025. Its S-1 shows actual annual revenue rising from about $41 million in 2023 to about $4.5 billion in 2025, as summarized by YourStory.

Run-rate milestones, as reported:

• Start of 2025: about $1 billion
• August 2025: more than $5 billion (Series F announcement)
• February 2026: $14 billion (Series G announcement)
• May 2026: more than $47 billion (Series H)
• End of July 2026: more than $65 billion (Bloomberg, August 2026, citing an investor update), up more than sevenfold from the end of 2025

The S-1 filed on October 26, 2026, gives booked numbers, as summarized by The VC Corner from the filing:

• First-half 2026 revenue: $16.29 billion, versus $1.16 billion in the first half of 2025
• Q2 2026 revenue: $11.56 billion, up about 1,368% year over year
• Customers paying more than $100,000 a year (trailing twelve months): from more than 950 to more than 11,900 between September 30, 2025, and September 30, 2026
• Net revenue retention: roughly 900% as of September 30, 2026, meaning the average existing customer cohort spent about nine times more than a year earlier

Run-rate comparison chart ($ billions, as reported)

Each block is about $2.5 billion.

OpenAI
End 2023   ~$2B   █
End 2024   ~$6B   ██
End 2025   $20B+  ████████
Aug 2026   $40B+  ████████████████
Sep 2026   ~$50B  ████████████████████   (FT)

Anthropic
Start 2025  ~$1B  ▌
Aug 2025    $5B+  ██
Feb 2026    $14B  ██████
May 2026    $47B+ ███████████████████
Jul 2026    $65B+ ██████████████████████████

Two cautions apply. First, the companies define run-rate differently, and Bloomberg has noted that "differences in accounting methodologies between the private startups make direct comparisons complex." Second, run-rate shows momentum, not cash in the bank. A company with a smaller run-rate but a giant, cheap-to-serve consumer audience can still be the stronger business.

Revenue mix: consumers vs enterprise vs coding

The two companies make their money in very different ways.

OpenAI started as a consumer subscription business. As recently as late 2024, reporting said most of its revenue came from ChatGPT subscriptions. By April 2026, OpenAI said enterprise customers made up about 40% of revenue, with more than 900 million weekly ChatGPT users and over 50 million paying subscribers. By August 2026 ChatGPT had about 1 billion weekly active users, and the advertising business OpenAI started testing in February 2026 had reached a $1 billion annualized run-rate in roughly 200 days (CNBC). OpenAI's own long-range plan, as reported by The Information in February 2026, targets about $150 billion of consumer revenue by 2030, $70 billion from ChatGPT Enterprise and business products, $47.5 billion from the API, and $15 billion from hardware and new devices.

Anthropic is mostly a business-to-business company. The Wall Street Journal reported in late 2025 that about 80% of its revenue came from business clients. Its fastest-growing product has been Claude Code, its coding agent. Claude Code's run-rate passed $2.5 billion by February 2026, more than doubling since January 1, and more than half of that came from enterprise customers. Menlo Ventures data cited by OfficeChai put Anthropic's share of enterprise large-language-model API usage at 40% in 2025, versus 27% for OpenAI and 21% for Google. In coding specifically, the figures were 54% for Anthropic and 21% for OpenAI. Anthropic has no advertising business. It ran a Super Bowl campaign in February 2026 built around keeping Claude ad-free.

The mix matters for a downturn. Enterprise and developer spending tends to be stickier once it is built into workflows, but it is also concentrated. If a few hundred large customers cut their AI budgets at once, the effect could be sharp. Consumer subscriptions are spread across tens of millions of people, but individual users cancel easily, and free users only pay off if ads or commerce work at scale.

Cash burn, losses and profitability timelines

This is where the two companies diverge most, at least on their own projections.

OpenAI: projections that keep getting bigger

OpenAI's internal cash-flow forecasts, as reported by The Information, have been revised upward several times:

Source: The Information, via The Decoder (September 2025 and February 2026). Negative figures are cash burn.

The September 2026 FT report went further. It put negative free cash flow at $278 billion from 2026 through 2030 and compute and infrastructure spending at about $856 billion by the end of 2030. It said OpenAI was on track to exhaust its current cash by 2028, which would require more fundraising. Revenue in that plan rises to about $350 billion in 2030, and cumulative revenue through 2030 comes to about $840 billion.

Margins have been a pressure point. The Information reported that OpenAI's adjusted gross margin fell to 33% in 2025, from 40% the year before and below a 46% internal target, after inference costs quadrupled. Its long-range gross-margin projections of 52% to 67% fell short of an earlier goal of 70% by 2029. Ed Zitron's reporting, which the FT said it verified, put OpenAI's losses at $5.09 billion in 2024 and $38.5 billion in 2025.

The Wall Street Journal reported in November 2025 that OpenAI projected a $74 billion loss in 2028 alone and did not expect to turn a profit until 2030. Sam Altman told reporters in 2025 that becoming profitable was "not in my top-10 concerns," according to The Decoder.

Anthropic: an earlier target, now partly visible in its filings

In late 2025, Anthropic's investor projections, reported by The Information and summarized by TechCrunch, called for:

• Gross margin rising from about 50% in 2025 to 77% in 2028, after a negative 94% the year before
• Up to $70 billion of revenue and $17 billion of positive cash flow in 2028
• Break-even in the 2027 to 2028 window, with outlets differing on the exact year

The S-1 gives the first audited-style look. As summarized from the filing:

• First-half 2026 net loss: $80.75 billion
• Of which non-cash fair-value changes on convertible notes, derivative liabilities and SAFEs: $72.39 billion. These charges rose as the company's valuation climbed, and no cash left the company for them.
• First-half 2026 operating loss: about $693 million, down from $2.14 billion a year earlier, per press coverage of the filing
• Q2 2026 operating income: $974 million, the first operating-profit quarter of the six shown
• Cash and short-term investments: about $131.7 billion

Read carefully, this shows a company whose core operations were roughly at break-even by mid-2026. The headline loss mostly reflects accounting for instruments whose value rose with the stock price. That does not make Anthropic safe. One profitable quarter during a period of explosive demand says little about how the business would hold up if prices or volumes fell. But the gap with OpenAI's reported trajectory is real.

Burn and profit targets side by side

Compute commitments: who owes what to whom

Compute is the fixed cost that would hurt most in a downturn. Training and serving frontier models takes data centers measured in gigawatts, and these contracts often run five to ten years.

OpenAI's web of deals

Totals vary by who is counting. The Decoder tallied about $1 trillion across more than 20 GW in October 2025, using OpenAI's own estimate of roughly $50 billion per gigawatt. The Wall Street Journal, as cited by Gizmodo, reported data center commitments of more than $1.4 trillion over eight years. Built In described more than $1 trillion of hardware and cloud agreements over ten years, with cloud commitments alone approaching $600 billion.

Critics have pointed out how circular some of these arrangements are. Suppliers invest in OpenAI, and OpenAI spends that money on the suppliers' products. Oracle's market value rose by about $244 billion after its deal was announced. "OpenAI is in no position to make any of these commitments," DA Davidson analyst Gil Luria said in October 2025.

Anthropic's multi-cloud approach

Anthropic has spread its compute across three chip families: Amazon's Trainium, Google's TPUs and Nvidia's GPUs.

• Project Rainier (AWS): Amazon activated this cluster of nearly 500,000 Trainium2 chips on October 29, 2025, less than a year after announcing it. AWS said it expected Anthropic to scale to more than one million Trainium2 chips by the end of 2025. An Amazon spokesperson said in October 2025 that "AWS continues to be both Anthropic's primary training partner and cloud provider."
• Google TPUs: Announced October 23, 2025. Anthropic gets access to up to one million TPUs and "well over a gigawatt" of capacity by 2026. Terms were not disclosed, but the deal was described as worth tens of billions of dollars.
• Microsoft and Nvidia: In November 2025 Anthropic committed to buy about $30 billion of Azure compute and to contract up to 1 GW of Nvidia Grace Blackwell and Vera Rubin systems.
• 2026 expansions: Around the Series H, Anthropic cited up to 5 GW of new capacity from Amazon, 5 GW of TPU capacity from Google and Broadcom, and GPU capacity in SpaceX's Colossus 1 and Colossus 2 data centers.
• S-1 total: Named contracts with Google, Amazon, Microsoft, SpaceXAI, Broadcom, Nscale and Lambda add up to roughly $637 billion, according to The VC Corner's sum of the filing's pages.

The multi-cloud approach gives Anthropic negotiating leverage and some resilience. Claude stayed online during a major AWS outage in 2025, for example. But it also means Anthropic depends on Amazon and Google, two companies that build competing models. And $637 billion is not small. It is less than OpenAI's headline totals, but it is still a commitment that only works if demand keeps growing quickly.

Commitment scale chart (headline totals, $ billions)

These totals use different methods and different time frames, so read them as rough orders of magnitude. Each block is about $50 billion.

OpenAI (WSJ, "more than", ~8 yrs)  $1,400B+ ████████████████████████████
OpenAI (The Decoder, Oct 2025)     ~$1,000B ████████████████████
Anthropic (S-1 named contracts)    ~$637B   █████████████

Structure and control: OpenAI's PBC vs Anthropic's Long-Term Benefit Trust

OpenAI: from capped profit to public benefit corporation

OpenAI began in 2015 as a nonprofit, later added a "capped-profit" arm, and on October 28, 2025, completed a recapitalization. The for-profit company became OpenAI Group PBC, a public benefit corporation. The nonprofit, renamed the OpenAI Foundation, controls it and appoints its board. Microsoft received about 27% of the PBC, valued at roughly $135 billion at the time, on its $13.8 billion of cumulative investment. Under the new terms:

• Microsoft keeps certain rights to OpenAI's models and products until at least 2032, even if OpenAI declares artificial general intelligence (AGI).
• An independent expert panel must verify any AGI claim, which matters because many provisions of the 2019 agreement depended on that milestone.
• OpenAI committed to buy $250 billion of Azure services, and Microsoft lost its right of first refusal on OpenAI's compute.

"The nonprofit remains in control of the for-profit, and now has a direct path to major resources before AGI arrives," Foundation chair Bret Taylor said, according to Al Jazeera.

Anthropic: a PBC with a trust on top

Anthropic was set up as a public benefit corporation. In September 2023 it announced a Long-Term Benefit Trust (LTBT), "an independent body of five financially disinterested members." The Trust holds a special class of stock, Class T, that "grants the Trust the authority to elect and remove a number of Anthropic's board members," and its power phases in over time. Anthropic said the Trust "will elect a majority of the board within 4 years." Large stockholder supermajorities can amend the arrangement, and "the required supermajorities increase as the Trust's power phases in." Trustees have changed over the years. Ben Bernanke, the former Federal Reserve chair, joined in July 2026.

The S-1 adds a twist for public investors. According to The VC Corner's reading of the filing, the seven co-founders hold 50.1% of the vote through a single Class F share held by a founder LLC, while the LTBT elects a majority of the board. Listed shares are Class A. In other words, public buyers would get economic exposure with limited control, which is common for founder-led tech IPOs.

What structure means in a downturn

In a crunch, governance decides who can force hard choices. OpenAI's nonprofit control and Microsoft's large stake mean any rescue or restructuring would involve a powerful incumbent partner with its own interests. Anthropic's founder vote and trust would make a hostile takeover very hard. They could also make it harder for public shareholders to push for cost cuts if growth stalls.

IPO talk

• Anthropic reportedly filed confidentially earlier in 2026, chose Nasdaq, and filed a public S-1 on October 26, 2026, under the ticker ANT. Goldman Sachs and Morgan Stanley lead the deal. In September 2026, The Next Web cited reports of a raise of up to $100 billion, which would be the largest IPO ever. The filing left the price range, share count and raise size blank. Timing reports have shifted from October to a window after the November 3 US midterm elections and before Thanksgiving. As of October 31, 2026, no shares had been priced.
• OpenAI reportedly filed confidentially in June 2026. But on September 12, 2026, Sam Altman told Fortune that OpenAI would not list in 2026, calling it an "ill-advised moment to go public," according to The Next Web. OpenAI has since signaled a listing in 2027.

The order matters. If Anthropic lists first at a premium price, it gives public investors a benchmark for valuing frontier AI labs, and OpenAI will be judged against it. If the IPO struggles, it will be read as an early sign that the bubble is losing air.

Key people and departures

OpenAI

OpenAI has lost a long list of senior figures since 2024:

• Ilya Sutskever, co-founder and chief scientist, left in 2024 after the November 2023 boardroom conflict in which Altman was briefly fired and then reinstated.
• John Schulman, a co-founder, left for Anthropic in August 2024.
• Mira Murati, chief technology officer, left in September 2024, along with research VP Barret Zoph and others.
• In 2026, Kevin Weil (OpenAI for Science), Bill Peebles (head of Sora) and Srinivas Narayanan (CTO of enterprise applications) announced their departures on April 17, as OpenAI shut down "side quests" including the Sora video app. The Next Web reported that only two of OpenAI's 11 co-founders remained: Altman and Greg Brockman.

Anthropic

Anthropic's seven co-founders still control 50.1% of the vote, according to the S-1, and the company has not had a wave of executive exits comparable to OpenAI's. It has, however, had public safety-related resignations:

• Mrinank Sharma, who led safeguards research, resigned in February 2026 with a letter warning that the world was "in peril."
• Jacob Coxon, a researcher who had spent three years training models, first at OpenAI and then at Anthropic, resigned in September 2026. He said in posts on X that both companies were "gambling with our lives" and that "neither company is acting responsibly," according to NPR member station KUNC.

Departures like these hurt Anthropic's brand more than OpenAI's, because Anthropic's pitch rests so heavily on being the careful lab.

Dario Amodei vs Sam Altman

The rivalry between the two companies is also a rivalry between two very different chief executives who once worked together.

Backgrounds

Sam Altman was born in Chicago in 1985. He studied computer science at Stanford for two years and dropped out in 2005 to co-found Loopt, a location-sharing app that raised over $30 million and was sold to Green Dot in 2012 for $43.4 million. He became a partner at the startup accelerator Y Combinator in 2011 and its president in 2014, succeeding Paul Graham. He co-founded OpenAI as a nonprofit in 2015 with Greg Brockman, Elon Musk and others, became CEO in 2019, and oversaw the launch of ChatGPT in November 2022. In November 2023 OpenAI's board fired him, saying he had not been "consistently candid." After an employee and investor revolt, he was reinstated five days later with a new board.

Dario Amodei was born in San Francisco in 1983. He studied at Caltech, earned a physics degree at Stanford, and earned a PhD in biophysics at Princeton, where he studied neural circuits. He worked at Baidu and Google Brain before joining OpenAI in 2016. As vice president of research, he helped lead work on GPT-2, GPT-3 and reinforcement learning from human feedback. According to Wikipedia's summary of reporting, when OpenAI's board ousted Altman in November 2023, it approached Amodei about replacing him and possibly merging the two companies. He declined.

The 2021 split

Dario Amodei, his sister Daniela Amodei and several colleagues left OpenAI at the end of 2020 and founded Anthropic in 2021. Neither side has given a full public account. In a 2023 interview with Fortune, Amodei described two convictions the founding group shared: first, "the idea that if you pour more compute into these models, they'll get better and better," and second, that "you needed something in addition to just scaling the models up, which is alignment or safety." He added, "We really trusted each other and wanted to work together. And so we went off and started our own company with that idea in mind." Other accounts describe disagreements over safety and direction, and over OpenAI's commercial turn after Microsoft's investment. Those accounts are secondhand, and the founders have generally avoided specific public criticism of their former employer.

Public jabs and rare agreement, 2026

• The Super Bowl ads (February 2026): Weeks after OpenAI confirmed it would test ads in ChatGPT's free and lower-cost tiers, Anthropic ran a campaign mocking ad-supported AI. The spots showed a personified algorithm interrupting serious conversations to pitch payday loans and height-boosting insoles, and positioned Claude as ad-free. Before the game, on February 4, Altman posted a long response on X. He called the ads funny but "clearly dishonest" and accused Anthropic of "doublespeak." Anthropic declined to comment to reporters.
• The Pentagon split (February 2026): Anthropic refused to drop contract terms barring the use of Claude for mass domestic surveillance or fully autonomous weapons. On February 27, 2026, the administration ordered agencies to stop using Anthropic's tools, and the Defense Department designated Anthropic a "supply chain risk," the first such designation of an American company, according to Fortune. Its $200 million Pentagon contract was cancelled. Hours later, OpenAI announced a deal for classified use of its models. Altman said the agreement included the same two limits Anthropic had sought, and wrote, "We are asking the [Department of War] to offer these same terms to all AI companies." Numerous OpenAI employees signed an open letter backing Amodei's position. In March 2026, a federal judge temporarily blocked the Defense Department's action.
• The handholding moment (February 19, 2026): At the India AI Impact Summit in New Delhi, Prime Minister Narendra Modi invited the tech leaders beside him to link raised hands for a photo. Altman and Amodei, standing next to each other, ended up raising clenched fists instead, and the image went viral. Altman later said, "I was sort of confused and didn't know what I was supposed to do."
• "Pace the frontier" (September 2026): Amodei published an essay arguing that leading labs should leave more time between major capability jumps, invite independent evaluators with deep access, coordinate across the industry and eventually reach international agreements, including with China. On September 12, Altman replied on X: "I agree with Dario that we need to pace the frontier," and "Committing to having independent evaluators with employee-like access is a great idea, and we will do the same." Elon Musk also posted support. President Donald Trump rejected the call to slow down.

Differing views

• Risk: Amodei has put the chance of catastrophic harm from AI at 10% to 25%, and his January 2026 essay The Adolescence of Technology lists five categories of risk, from misaligned autonomous systems to misuse for mass destruction. Altman has also talked about safety since OpenAI's founding, but in public he leans more toward deployment. In 2025 he wrote that "the rate of technological progress will keep accelerating." In September 2026 he told a G20 meeting that adopting AI was "non-negotiable."
• Jobs: In May 2025, Amodei warned that AI could eliminate half of all entry-level white-collar jobs within one to five years and push unemployment to 10% to 20%. Altman wrote in 2025 that "there will be very hard parts like whole classes of jobs going away," but argued that "the world will be getting so much richer so quickly" that new policies would become possible. He has proposed basic-income ideas, but said in April 2026 that he believes in universal basic income less than he once did.
• Regulation: Altman testified to a Senate Judiciary subcommittee on AI oversight in May 2023. Amodei has called for industry and international coordination, and wrote in September 2026 that "some forms of coordination that would be impactful for pacing are legally challenging, and will require government support."
• China: Amodei is the more hawkish of the two. He supports strict semiconductor export controls and safeguards against model distillation, and has argued that democracies must keep an AI lead. In Machines of Loving Grace (October 2024) he wrote that "AI-powered authoritarianism seems too terrible to contemplate." Altman has stressed dialogue between US and Chinese leaders while acknowledging that approval for cooperation is uncertain.

Safety vs speed: positioning, and how it plays in a downturn

Anthropic has built its brand around caution: the Long-Term Benefit Trust, the published essays, the Pentagon refusal and the ad-free promise. OpenAI has built its brand around reach and speed: a consumer superapp, a hardware push, ads, commerce and the biggest infrastructure build-out in tech.

In a bubble deflation, each position cuts both ways. The safety brand helps Anthropic with risk-averse enterprises, regulators and governments outside the US, and it could matter more if a high-profile AI failure drives a backlash. But it also cost Anthropic US federal business in 2026, and resignations like Sharma's and Coxon's show that even the "careful" lab draws criticism for moving too fast. OpenAI's speed has won it an enormous consumer audience and a broad product line. The same speed, though, led to products it later shut down, such as Sora, which cost roughly $1 million a day to run and peaked at about one million users before falling below 500,000, according to The Next Web. When money gets tight, products like that are the first to go.

Nvidia: the arms dealer to both labs

Whatever happens between OpenAI and Anthropic, one company has profited from both: Nvidia. Each lab also uses alternatives. OpenAI has deals with AMD and Broadcom, and Anthropic uses Trainium and TPUs. But Nvidia GPUs remain the default currency of AI compute, and Nvidia has invested directly in both labs.

• OpenAI: On September 22, 2025, Nvidia and OpenAI announced a memorandum of understanding for up to $100 billion of investment tied to at least 10 GW of Nvidia systems. By January 2026, negotiations had not moved past early stages, and Nvidia CEO Jensen Huang had privately described the original deal as non-binding. In the round OpenAI closed in March 2026, Nvidia committed $30 billion, and OpenAI pledged 2 GW of Vera Rubin training capacity plus 3 GW of inference compute.
• Anthropic: In November 2025, Nvidia committed up to $10 billion, alongside up to $5 billion from Microsoft, and Anthropic agreed to contract up to 1 GW of Grace Blackwell and Vera Rubin systems. Nvidia also took part in Anthropic's February 2026 Series G.

This is the "circularity" problem in its plainest form. Nvidia invests in its customers, and those customers spend the money on Nvidia hardware. That supports demand in a boom. In a bust, it could magnify losses on both sides.

Data center vs gaming revenue

Nvidia's fiscal year ends in late January, so fiscal 2026 mostly covers calendar 2025.

\FY2023 segment figures are approximate, back-calculated from the growth rates in Nvidia's FY2024 release. \\*Starting with this quarter, Nvidia stopped breaking out gaming. The remaining $7.2 billion is reported as "Edge Computing," which bundles gaming with other products.

Nvidia annual revenue by segment ($B), each block is about $10B

FY2023  Data Center ██           ~15
        Gaming      █            ~9
FY2024  Data Center █████        47.5
        Gaming      █            10.4
FY2025  Data Center ████████████ 115.2
        Gaming      █            11.4
FY2026  Data Center ███████████████████ 193.7
        Gaming      ██           16.0

Gaming revenue did grow. FY2026's $16.0 billion was a record, up 41%, driven by Blackwell-based RTX 50-series cards. But gaming fell from roughly a third of Nvidia's business to a rounding error next to the data center. In Q2 FY2027, Nvidia earned $59.7 billion of net income in one quarter on a 75% gross margin, and guided to about $108 billion of revenue for the following quarter.

Market cap milestones

• $1 trillion: reached in 2023
• $2 trillion: first close above it on March 1, 2024
• World's most valuable company: June 18, 2024, at more than $3.3 trillion
• DeepSeek shock: January 2025, a one-day loss of about $600 billion in market value, the largest for any US company, after a Chinese startup showed a capable model built with far less compute
• $4 trillion: first close above it on July 10, 2025, the first company ever to do so
• $5 trillion: October 29, 2025, again a first

Reports through 2026 have put Nvidia above $5 trillion at various points, but this article could not confirm a specific later milestone from a primary source.

What a bubble deflation would mean for Nvidia

Nvidia is the most direct public-market bet on AI capital spending continuing. If OpenAI, Anthropic and the hyperscalers that serve them slow their build-outs, the first effect would be on Nvidia's data center orders, which made up about 92.5% of revenue in its latest quarter. Several specific risks stand out:

• Customer concentration: Much of the demand traces back to a handful of hyperscalers and AI labs. Nvidia has invested in two of the biggest.
• Circular exposure: Equity stakes in OpenAI and Anthropic would lose value at the same moment orders weakened.
• Margins: A 75% gross margin depends on scarcity. In a glut, customers would gain pricing power, and alternatives such as TPUs, Trainium, AMD and custom Broadcom chips would compete harder.
• The DeepSeek precedent: January 2025 showed how quickly investors can mark down Nvidia on a story about cheaper AI, even before order books change.

The counterargument is that Nvidia is not a startup. It is enormously profitable, returned about $26 billion to shareholders in Q2 FY2027 alone, and would survive a downturn with a strong balance sheet. A deflation would probably mean a sharp fall in growth and valuation, not a threat to the company's survival.

RTX implications: what the AI boom means for gamers and PC buyers

For ordinary consumers, the AI race shows up in GPU prices, memory prices and availability.

The RTX 50-series rollout. Nvidia launched the RTX 5090 ($1,999) and RTX 5080 ($999) on January 30, 2025. The RTX 5070 Ti ($749), RTX 5070 ($549), RTX 5060 Ti ($379 for 8GB, $429 for 16GB) and RTX 5060 ($299) followed between February and May 2025. Launch supply was badly short. Tom's Guide reported a UK retailer warning of four-month waits for the 5090 and 5080. VideoCardz reported board partners' own stores pricing the 5090 at up to $3,409. Reviewers said the 5070 Ti was usually available only at $900 or more, and TechPowerUp called some MSRPs "pure fiction."

The memory crunch, 2025 to 2026. The bigger shock came from memory. Chipmakers shifted wafer capacity toward high-bandwidth memory (HBM) for AI accelerators, squeezing supply of ordinary DDR4 and DDR5 for PCs and phones. Reports said OpenAI's Stargate letters of intent with Samsung and SK Hynix in October 2025 could absorb up to about 900,000 DRAM wafers a month, around 40% of global output. The effects, as compiled from Reuters, TrendForce, Tom's Hardware, CNBC and others:

• DRAM prices reportedly rose 172% in 2025. A former Samsung executive cited a 414% spike in DDR5 prices.
• November 2025 NAND wafer contract prices rose more than 60% in a single month.
• Dell raised PC prices 15% to 20% in December 2025. HP said memory made up 35% of a PC's bill of materials in Q1 2026. Apple raised prices on Macs and iPads in June 2026.
• Micron announced it was exiting its Crucial consumer memory business.
• DDR5 prices eased nearly 30% by April 2026 but stayed far above normal.
• Gartner and IDC expected PC shipments to fall 10% to 11% in 2026. Gartner warned that laptops under $500 could become financially unviable within two years.

The squeeze reached graphics cards too. In January 2026, TechRadar warned of possible RTX 50-series shortages caused by the RAM crisis, and Tom's Hardware described an AI-driven pricing crisis. Reuters reported that Nvidia acknowledged significant price increases but said it had secured adequate supply. In its fiscal 2026 results, Nvidia said gaming fell 13% from the previous quarter as channel inventory moderated after the holidays. Coverage of CFO Colette Kress's commentary quoted her saying, "We expect supply constraints in the first quarter of fiscal year 2027 and beyond to be a headwind for gaming."

RTX 5090 discontinuation reports. Starting around October 10, 2026, outlets including Tom's Hardware, TechPowerUp and TechSpot reported that the RTX 5090 was being discontinued and that its GB202 chips were being diverted to server and workstation products. Micro Center reportedly stopped carrying it, and it disappeared from Nvidia's online store. Nvidia denied the discontinuation, according to TweakTown on October 17, 2026. Tom's Hardware questioned the denial and noted a 5090 listing at $8,500 on Amazon. As of this writing, the card's long-term status is unresolved. Rumors of a 24GB RTX 5080 as a replacement flagship remain unconfirmed.

Local AI on RTX PCs. One side effect of the boom is that consumer GPUs are now also sold as personal AI machines. Running open-weight models locally rewards cards with more video memory, which adds a second group of buyers competing with gamers for high-memory cards. Nvidia released an "RTX Spark" chip for laptops on June 1, 2026, according to Wikipedia's summary of its 2026 announcements.

If the bubble pops, do gamers win? Some things are reasonable to expect, but they are analysis, not reported fact:

• Memory prices are the most likely relief. If AI build-outs slow, wafer capacity pulled into HBM could swing back to commodity DRAM and GDDR. Industry forecasts differ widely even without a bust. Micron's CEO said in June 2026 that the shortage would last through 2027 and supply would improve gradually by 2028. SK Hynix's chief said in August 2026 it could persist until 2030. Acer's CEO predicted in September 2026 that prices would start falling in late 2027, and accused memory makers of prolonging shortage forecasts to protect profits. A sharp slowdown in AI spending would probably pull those dates forward.
• GPU prices would likely ease, but not collapse. Nvidia makes far more money from data center chips than from gaming chips, so an AI slowdown would leave it more willing to send scarce wafers and memory back to GeForce. That should improve supply and narrow the markup over MSRP. Whether Nvidia would actually cut list prices is a separate question that depends on AMD and Intel competition.
• Don't expect a flood of cheap data-center cards for gaming. Many top AI accelerators, such as SXM-format H100 and Blackwell modules, have no display outputs. They need specialized servers, power and cooling, and they are not designed to run games. A wave of used data-center hardware would mostly help small AI labs, universities and hobbyists running local models, not people buying a card for gaming.

Bull and bear cases if the bubble deflates

OpenAI

Bull case

• About 1 billion weekly ChatGPT users, the largest audience in AI, gives OpenAI the most options to make money: subscriptions, ads (already a $1 billion run-rate within roughly 200 days), commerce and devices.
• It raised $122 billion in early 2026 and held more than $73 billion of cash after Q1. Its backers, including Microsoft, Amazon, Nvidia and SoftBank, all have strategic reasons to keep it alive.
• Its compute deals are spread across nearly every major supplier. In a glut, OpenAI could renegotiate from strength because its suppliers need it as a customer.
• Its brand is close to synonymous with AI for consumers.

Bear case

• Its own projections show about $278 billion of cash burn from 2026 to 2030 and cash running out by 2028, according to the FT. A downturn that closes funding markets would hit exactly when its needs peak.
• Its compute commitments, tallied at $1 trillion to $1.4 trillion or more, are fixed costs set against uncertain revenue.
• Gross margin fell to 33% in 2025. Free users are expensive to serve, and ads have to work at enormous scale to cover them.
• Repeated leadership departures and abandoned projects like Sora suggest a company stretched across too many bets.
• It ruled out a 2026 IPO, so it remains dependent on private capital.

Anthropic

Bull case

• It reported operating income of $974 million in Q2 2026 and about $131.7 billion in cash and short-term investments, a much larger cushion relative to its burn.
• Its revenue comes mostly from enterprise and developer work, especially coding, where AI spending is tied to measurable productivity and where it led API and coding usage in 2025.
• Its compute is spread across Trainium, TPUs and Nvidia GPUs, so it is not locked into one supplier's pricing.
• A public listing would give it access to public capital markets that OpenAI has put off until at least 2027.

Bear case

• An IPO priced near the reported $2 trillion would leave no room for slowing growth. Net revenue retention of about 900% cannot last, and growth will slow even without a bust.
• Named compute contracts of about $637 billion are large fixed obligations.
• It relies on Amazon and Google, which are investors and suppliers but also rivals with their own models.
• Concentration in enterprise and coding means a corporate budget freeze would hit fast. Price competition in coding agents, including from cheaper open-weight models, could squeeze margins.
• Political risk is real. The 2026 Pentagon designation showed that a US administration could shut it out of federal business.
• Its large GAAP losses, even if mostly non-cash, will be the first thing skeptics point to.

Scorecard: side by side

On these figures, Anthropic looks better positioned for a short, sharp downturn because it is closer to self-funding and holds more cash. OpenAI looks better positioned for a long, grinding contest for consumers because of its audience and broader set of revenue sources. Neither is cheap, and the scorecard could change quickly.

Timeline

• 2015: OpenAI founded as a nonprofit.
• 2016: Dario Amodei joins OpenAI.
• End of 2020 to 2021: Dario and Daniela Amodei and colleagues leave OpenAI and found Anthropic.
• November 2022: ChatGPT launches.
• 2023: Anthropic's run-rate reaches about $100 million. Nvidia's market value reaches $1 trillion.
• September 19, 2023: Anthropic announces the Long-Term Benefit Trust.
• November 2023: OpenAI's board fires Altman, then reinstates him five days later.
• March 1, 2024: Nvidia closes above $2 trillion.
• August to September 2024: John Schulman leaves OpenAI for Anthropic. Mira Murati leaves OpenAI.
• October 2, 2024: OpenAI raises $6.6 billion at $157 billion.
• October 2024: Amodei publishes Machines of Loving Grace.
• January 2025: Stargate is announced. DeepSeek wipes about $600 billion off Nvidia's value in one day.
• January to May 2025: The RTX 50 series launches amid shortages.
• March 2025: Anthropic raises $3.5 billion at $61.5 billion.
• March 31, 2025: OpenAI raises up to $40 billion at $300 billion, led by SoftBank.
• May 2025: Amodei warns about entry-level white-collar jobs.
• July 10, 2025: Nvidia closes above $4 trillion.
• September 2, 2025: Anthropic raises $13 billion at $183 billion.
• September 22, 2025: Nvidia and OpenAI sign a $100 billion letter of intent.
• October 2025: OpenAI's AMD and Broadcom deals. Samsung and SK Hynix memory letters of intent. Anthropic's Google TPU deal on October 23. Project Rainier goes live on October 29. OpenAI's PBC recapitalization on October 28. Nvidia reaches $5 trillion on October 29.
• November 18, 2025: Microsoft and Nvidia commit up to $15 billion to Anthropic.
• Late 2025: The memory shortage drives PC price increases.
• January 2026: Amodei publishes The Adolescence of Technology.
• February 2026: ChatGPT ads begin testing. Anthropic's Super Bowl ads air, and Altman calls them "clearly dishonest." Mrinank Sharma resigns. Anthropic raises $30 billion at $380 billion on February 12. Altman and Amodei have their awkward photo moment in New Delhi on February 19. The Pentagon designates Anthropic a supply-chain risk and OpenAI signs a Pentagon deal on February 27.
• March 31, 2026: OpenAI closes $122 billion at $852 billion.
• April 17, 2026: Weil, Peebles and Narayanan announce they are leaving OpenAI. Sora shuts down.
• May 28, 2026: Anthropic raises $65 billion at $965 billion.
• June 2026: Both companies reportedly file confidentially for IPOs.
• Late July 2026: Anthropic's run-rate passes $65 billion.
• August 2026: OpenAI's run-rate tops $40 billion. ChatGPT ads reach a $1 billion run-rate. Nvidia reports $96.2 billion of quarterly revenue and stops breaking out gaming.
• September 2026: The FT reports OpenAI's $278 billion burn projection. Altman rules out a 2026 IPO. Amodei's "pace the frontier" proposal wins Altman's endorsement. Jacob Coxon resigns from Anthropic.
• October 8, 2026: The FT puts OpenAI's September run-rate at about $50 billion.
• October 10 to 17, 2026: Reports say the RTX 5090 is being discontinued, and Nvidia denies it.
• October 26, 2026: Anthropic files its public S-1 for Nasdaq under the ticker ANT.

FAQ

Is Anthropic now bigger than OpenAI? By last private valuation ($965 billion vs $852 billion) and by reported run-rate ($65 billion+ vs about $50 billion), yes, as of late 2026. By users, no. ChatGPT had about 1 billion weekly users. The two companies measure run-rate differently, so the revenue comparison is less exact than it looks.

Is either company profitable? Not on a GAAP net basis. Anthropic reported its first operating-profit quarter in Q2 2026 ($974 million), but a first-half net loss of $80.7 billion, mostly from non-cash fair-value charges. OpenAI's projections show heavy cash burn until it turns cash-flow positive around 2030.

Why was Anthropic's loss so large if it made an operating profit? Most of the $80.7 billion was a $72.4 billion non-cash charge for changes in the fair value of convertible notes, derivatives and SAFEs. Those instruments became more expensive on paper as Anthropic's valuation soared. No cash left the company for that line.

How much does OpenAI owe in compute? Estimates range from about $1 trillion (The Decoder, October 2025) to more than $1.4 trillion over eight years (WSJ, via Gizmodo). The FT's September 2026 report put compute and infrastructure spending at about $856 billion through 2030.

Who owns OpenAI now? OpenAI Group PBC is controlled by the nonprofit OpenAI Foundation. Microsoft held about 27% after the October 2025 recapitalization. SoftBank, Amazon, Nvidia, Thrive and many others hold stakes, and those stakes have been diluted further by the 2026 round.

What is Anthropic's Long-Term Benefit Trust? An independent body of five financially disinterested trustees that holds Class T stock and elects a growing share of Anthropic's board, eventually a majority. It is meant to make sure the company weighs its public-benefit mission alongside shareholder returns.

When will each company go public? Anthropic filed a public S-1 on October 26, 2026, with a listing reportedly targeted for November 2026. Price and size had not been set. OpenAI said it would not list in 2026, and has signaled 2027.

What happens to Nvidia if the AI bubble pops? Data center sales, about 92.5% of its latest quarterly revenue, would slow, margins would come under pressure, and its stakes in OpenAI and Anthropic would lose value. Nvidia is highly profitable and cash-rich, so the risk is mainly to its growth and valuation, not its survival.

Will GPUs get cheaper if AI spending slows? Probably, especially memory-heavy cards, as memory supply moves back from HBM. That is an expectation, not a certainty. Don't expect used data-center accelerators to replace gaming cards. Many have no display outputs and need server infrastructure.

Do Altman and Amodei agree on anything? Yes. In September 2026, Altman publicly endorsed Amodei's call to "pace the frontier" and committed OpenAI to independent evaluators with employee-like access.

Sources and further reading

• Built In, "Anthropic Raises $65B Series H, Achieves $965B Valuation" (June 1, 2026): https://builtin.com/articles/anthropic-raises-65b-series-h-965b-valuation-20260601
• Built In, "OpenAI Raises $122B in Funding at $852B Valuation" (April 1, 2026): https://builtin.com/articles/openai-raises-122b-852b-valuation-20260401
• Built In, "Anthropic Increases Valuation to $183B With $13B Series F Round" (September 2, 2025): https://builtin.com/articles/anthropic-raises-13b-series-f-183b-valuation-20250902
• Built In, "OpenAI Is Building Its Own Global AI Infrastructure One Cloud Deal at a Time": https://builtin.com/articles/openai-cloud-deals
• Anthropic, "Anthropic raises $30 billion Series G funding at $380 billion post-money valuation" (February 12, 2026): https://www.anthropic.com/news/anthropic-raises-30-billion-series-g-funding-380-billion-post-money-valuation
• Anthropic, "The Long-Term Benefit Trust" (September 19, 2023): https://www.anthropic.com/news/the-long-term-benefit-trust
• Bloomberg Law, "Anthropic Finalizes $30 Billion Funding at $380 Billion Value" (February 12, 2026): https://news.bloomberglaw.com/financial-accounting/anthropic-finalizes-30-billion-funding-at-380-billion-value
• Bloomberg Government, "Anthropic's Annualized Revenue Tops $65 Billion Before IPO" (August 18, 2026): https://news.bgov.com/legal-ops-and-tech/anthropic-revenue-run-rate-surpasses-65-billion-ahead-of-ipo
• The Star (Reuters/FT), "OpenAI forecasts cash burn near $280 billion by 2030, FT reports" (September 19, 2026): https://www.thestar.com.my/tech/tech-news/2026/09/19/openai-forecasts-cash-burn-near-280-billion-by-2030-ft-reports
• The Next Web, "Anthropic has chosen Nasdaq for its October IPO, in the week OpenAI ruled one out" (September 13, 2026): https://thenextweb.com/news/anthropic-ipo-nasdaq-october-listing
• The Next Web, "Three more senior executives leave OpenAI as the company kills its side quests" (April 2026): https://thenextweb.com/news/openai-departures-kevin-weil-sora-peebles-enterprise-pivot
• YourStory, "Anthropic files for IPO even as losses surge to $80B" (October 26, 2026): https://yourstory.com/2026/10/anthropic-files-for-ipo-even-as-losses-surge-to-80-billion
• The VC Corner, "Anthropic S-1: IPO Filing, ANT Ticker, Revenue and Losses": https://www.thevccorner.com/p/anthropic-s1-ipo-filing
• OfficeChai, "OpenAI's Revenue Touched $20 Billion In 2025, Says Company" (January 19, 2026): https://officechai.com/ai/openais-revenue-touched-20-billion-in-2025-says-company
• PYMNTS, "OpenAI Ran Through $3.7 Billion in Q1 2026" (June 17, 2026): https://www.pymnts.com/?p=3859081
• Yahoo Finance, "OpenAI's annualized revenue $20 billion lower than prior investor estimates" (October 8, 2026): https://finance.yahoo.com/technology/article/openais-annualized-revenue-20-billion-lower-than-prior-investor-estimates-174058048.html
• Yahoo Finance (Investing.com/Bloomberg), "OpenAI revenue run rate tops $40 billion, Bloomberg reports" (August 13, 2026): https://finance.yahoo.com/technology/ai/articles/openai-revenue-run-rate-tops-224009196.html
• CNBC, "OpenAI's ad business shows blistering growth, hits $1 billion annualized revenue run rate" (August 31, 2026): https://www.cnbc.com/2026/08/31/open-ai-chatgpt-ads-revenue.html
• Engadget, "OpenAI secures another $122 billion in funding from Amazon, NVIDIA, SoftBank and others": https://www.engadget.com/ai/openai-secures-another-110-billion-in-funding-from-amazon-nvidia-and-softbank-171006356.html
• Al Jazeera, "OpenAI restructures into public-benefit firm, Microsoft takes 27% stake" (October 28, 2025): https://www.aljazeera.com/economy/2025/10/28/openai-restructures-into-public-benefit-firm-microsoft-takes-27-stake
• Fortune, "OpenAI officially raises $6.6 billion" (October 2, 2024): https://fortune.com/2024/10/02/openai-officially-raises-6-6-billion-funding-deal-157-billion-valuation-sam-altman-thrive-capital/
• SiliconANGLE, OpenAI $40B round at $300B (March 31, 2025): https://siliconangle.com/?p=691422
• SiliconANGLE, "Anthropic strikes multibillion-dollar deal with Google for access to a million TPUs" (October 23, 2025): https://siliconangle.com/2025/10/23/anthropic-strikes-multibillion-dollar-deal-google-access-million-tpus/
• SiliconANGLE, "Sam Altman and Elon Musk back Dario Amodei's call to slow down the frontier of AI development" (September 13, 2026): https://siliconangle.com/2026/09/13/sam-altman-and-elon-musk-back-dario-amodeis-call-to-slow-down-the-frontier-of-ai-development/
• Outlook Business, "Anthropic to Draw $15 Bn from Microsoft, Nvidia After Committing $30 Bn to Azure, 1GW of Nvidia Compute" (November 19, 2025): https://www.outlookbusiness.com/artificial-intelligence/anthropic-to-draw-15-bn-from-microsoft-nvidia-after-committing-30-bn-to-azure-1gw-of-nvidia-compute
• The Decoder, "OpenAI has now signed $1 trillion worth of AI infrastructure contracts" (October 7, 2025): https://the-decoder.com/openai-has-now-signed-1-trillion-worth-of-ai-infrastructure-contracts/
• The Decoder, "OpenAI has reportedly misjudged its cash burn by $80 billion" (September 6, 2025): https://the-decoder.com/openai-has-reportedly-misjudged-its-cash-burn-by-80-billion/
• The Decoder, "OpenAI adds $111 billion to its cash burn forecast" (February 21, 2026): https://the-decoder.com/openai-adds-111-billion-to-its-cash-burn-forecast-as-ai-costs-spiral-beyond-projections/
• TechCrunch, "Anthropic expects B2B demand to boost revenue to $70B in 2028: report" (November 4, 2025): https://techcrunch.com/2025/11/04/anthropic-expects-b2b-demand-to-boost-revenue-to-70b-in-2028-report/
• Gizmodo, "OpenAI will lose $74 billion the same year that Anthropic breaks even: report" (November 11, 2025): https://gizmodo.com/openai-will-lose-74-billion-the-same-year-that-anthropic-breaks-even-report-2000684287
• Fortune, Dario Amodei interview on leaving OpenAI (September 26, 2023): https://www.fortune.com/2023/09/26/anthropic-ceo-interview-quit-open-ai-amazon-investment
• Fortune, "OpenAI Pentagon deal; Anthropic designated a supply chain risk" (February 28, 2026): https://fortune.com/2026/02/28/openai-pentagon-deal-anthropic-designated-supply-chain-risk-unprecedented-action-damage-its-growth
• iClarified, "Anthropic mocks ChatGPT ads in Super Bowl spot, Altman fires back" (February 5, 2026): https://www.iclarified.com/99846/anthropic-mocks-chatgpt-ads-in-super-bowl-spot-altman-fires-back-video
• Deccan Chronicle, "'Didn't Know What I Was Supposed To Do': Altman on Awkward Moment With Anthropic's Amodei" (February 19, 2026): https://www.deccanchronicle.com/nation/didnt-know-what-i-was-supposed-to-do-altman-on-awkward-moment-with-anthropics-amodei-1938406
• KUNC/NPR, "Anthropic researcher resigns amid AI safety concerns" (September 9, 2026): https://www.kunc.org/2026-09-09/anthropic-researcher-resigns-amid-ai-safety-concerns
• Sam Altman, "The Gentle Singularity" (2025): https://blog.samaltman.com/the-gentle-singularity
• Dario Amodei, "Machines of Loving Grace" (October 2024): https://www.darioamodei.com/essay/machines-of-loving-grace
• Wikipedia, "Dario Amodei": https://en.wikipedia.org/wiki/Dario_Amodei
• Wikipedia, "Sam Altman": https://en.wikipedia.org/wiki/Sam_Altman
• Wikipedia, "Nvidia": https://en.wikipedia.org/wiki/Nvidia
• Wikipedia, "GeForce RTX 50 series": https://en.wikipedia.org/wiki/GeForce_RTX_50_series
• Wikipedia, "2024–2026 global memory supply shortage": https://en.wikipedia.org/wiki/2024–2026_global_memory_supply_shortage
• NVIDIA, Q4 and fiscal 2026 results (SEC Form 8-K exhibit, February 25, 2026): https://www.sec.gov/Archives/edgar/data/1045810/000104581026000019/q4fy26pr.htm
• NVIDIA, "NVIDIA Announces Financial Results for Fourth Quarter and Fiscal 2025": https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-fourth-quarter-and-fiscal-2025
• NVIDIA, "NVIDIA Announces Financial Results for Fourth Quarter and Fiscal 2024": https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-fourth-quarter-and-fiscal-2024
• iPhone in Canada, "Nvidia Blows Past Expectations with $96.2 Billion Revenue in Q2 2027" (August 27, 2026): https://www.iphoneincanada.ca/2026/08/27/nvidia-blows-past-expectations-with-96-2-billion-revenue-in-q2-2027/
• Gagadget, "NVIDIA's Revenue Soars to $96.2 Billion in Q2, Data Centers Dominate with 92% Share" (August 27, 2026): https://gagadget.com/en/723508-nvidias-revenue-soars-to-962-billion-in-q2-data-centers-dominate-with-92-share/