OpenAI at $852B. Anthropic at $380B. Databricks at $134B. Over $1.3T in private valuations heading for public markets. Bubble or boom?
Get new posts delivered to your inbox. No spam, unsubscribe anytime.
In Q1 2026, a single quarter of venture funding shattered every record in history. 242 billion dollars flowed into AI companies -- 80% of all global venture capital. Four companies alone -- OpenAI, Anthropic, xAI, and Waymo -- raised $188 billion in three months. And now, the biggest of them are preparing to go public.
OpenAI at $852 billion. Anthropic at $380 billion. Databricks at $134 billion. Cerebras at $23 billion. Combined, they represent over $1.3 trillion in private valuations seeking public markets. If even half of them list by year-end, 2026 will produce the largest AI IPO wave in history -- dwarfing the dot-com class of 1999 in raw dollar terms.
The question isn't whether these IPOs will happen. It's whether the valuations make any sense.
Let me lay out the financials of the four biggest IPO candidates, because the details matter more than the headlines:
| Company | Last Valuation | ARR (Latest) | Profitable? | Expected IPO | Key Metric |
|---|---|---|---|---|---|
| OpenAI | $852B | $25B | No (2030 target) | Q4 2026 / 2027 | 910M weekly users |
| Anthropic | $380B | $19B | No (2027 target) | Q4 2026 | 19x revenue growth in 15 months |
| Databricks | $134B | $5.4B | FCF-positive | H2 2026 | 65% YoY growth |
| Cerebras | $23B | Undisclosed | No | April 2026 | $10B OpenAI compute deal |
And for context, one AI infrastructure company already went public:
CoreWeave (CRWV) IPO'd in March 2025 at $40/share, initially underwhelming -- the offering was downsized from $2.7B to $1.5B. Then the stock surged 250% by June. Then crashed 46% in November. It's now up roughly 200% from IPO price but with extreme volatility. Revenue hit $5 billion in 2025 -- the fastest any cloud company has reached that milestone -- with a $66.8 billion contracted backlog.
CoreWeave is the preview. The main event hasn't started yet.
On April 1, 2026, OpenAI closed a $122 billion funding round -- the largest private fundraise in history. Amazon led with $50 billion. Nvidia and SoftBank each contributed $30 billion. The post-money valuation: $852 billion.
To put that in perspective, that's more than the market cap of JPMorgan Chase. For a company that has never turned a profit.
The revenue story is real. OpenAI crossed $25 billion in annualized revenue by February 2026, generating roughly $2 billion per month. ChatGPT alone generated $8 billion in 2025, about 66% of total revenue. Enterprise seats grew 9x year-over-year, surpassing 9 million paying business users.
The loss story is also real. OpenAI projects a $14 billion loss in 2026, with ~$13 billion in revenue against ~$22 billion in spending. The cumulative deficit could reach $143 billion by 2029. Breakeven isn't expected until 2030. HSBC analysts have suggested OpenAI "likely won't make money by 2030" and faces a $207 billion funding shortfall.
The structural issues go deeper than the P&L. OpenAI's adjusted gross margin collapsed from 40% in 2024 to 33% in 2025 as inference costs quadrupled. ChatGPT's web traffic share fell from 86.7% in January 2025 to 64.5% in January 2026, while Google Gemini grew from 5.7% to 21.5%. The enterprise LLM API market tells the same story: OpenAI's share dropped from 50% in 2023 to roughly 25% by mid-2025.
And yet. 910 million weekly active users. $25 billion ARR. The most recognized AI brand on the planet. The IPO will be massive regardless of whether the valuation is justified.
OpenAI completed its conversion to a Delaware Public Benefit Corporation in October 2025, clearing the last regulatory hurdle. The nonprofit (now called the OpenAI Foundation) retains a 26% equity stake. CFO Sarah Friar has suggested 2027 may be more realistic than a 2026 listing, but analysts broadly expect the IPO between Q4 2026 and early 2027 at a potential valuation near $1 trillion.
Anthropic's trajectory is the most remarkable revenue story in tech history. The company went from $1 billion ARR in December 2024 to $19 billion ARR by March 2026 -- a 19x increase in 15 months. No enterprise software company has ever grown this fast at this scale.
In February 2026, Anthropic closed a $30 billion Series G at a $380 billion valuation -- the second-biggest private financing round on record (after OpenAI's). The company is now expected to surpass OpenAI in revenue by mid-2026.
Here's where Anthropic's story diverges from OpenAI's. The spending discipline is fundamentally different.
Anthropic plans to spend $19 billion in 2026 -- $12 billion on training, $7 billion on inference infrastructure. But the company projects dropping cash burn to one-third of revenue in 2026 and 9% by 2027, with positive free cash flow expected by 2027. Compare that to OpenAI's 2030 profitability target.
The business model difference is structural. Anthropic generates 80% of its revenue from enterprise customers, with eight Fortune 10 companies as clients. Its estimated net revenue retention rate is roughly 140% -- meaning existing customers spend 40% more each year. Anthropic hasn't launched a video generator, a robotics division, or a consumer hardware play. It sells API access and Claude subscriptions. That's it.
While OpenAI's enterprise API share fell from 50% to 25%, Anthropic's rose from 12% to 32% over the same period. The market is redistributing, and Anthropic is the primary beneficiary.
The IPO plan: S-1 filing expected late summer 2026, with a Nasdaq listing targeting Q4 2026 (potentially October). Bankers expect the IPO could raise more than $60 billion. Goldman Sachs, JPMorgan, and Morgan Stanley are in early talks.
Infrastructure-wise, Anthropic is expanding to up to 1 million Google Cloud TPUs -- over a gigawatt of capacity coming online in 2026. The compute strategy is diversified across Google TPUs, Amazon Trainium, and Nvidia GPUs. No single-vendor dependency.
Databricks gets less attention than OpenAI and Anthropic, but its IPO story might be the cleanest of the three.
The company raised a total of $9 billion across two rounds in late 2025 and early 2026, reaching a $134 billion valuation. Revenue hit $5.4 billion ARR in January 2026, up 65% year-over-year. AI products alone generate $1.4 billion.
The critical differentiator: Databricks is already free-cash-flow positive. No burn rate to explain away. No "profitability by 2030" projections. At 25x ARR on 65% growth with positive FCF, analysts at PitchBook call it "the cleanest institutional entry point in this cohort by every relevant measure."
CEO Ali Ghodsi said he "wouldn't rule out" an IPO this year. The company secured $1.8 billion in debt financing in January 2026 -- a classic pre-IPO signal. No S-1 filing as of early April, but H2 2026 is the consensus timeline.
Databricks doesn't make headlines like OpenAI. It doesn't have 910 million users or a billion-dollar consumer product. What it has is a data platform that enterprises actually pay for, revenue that's growing 65% annually, and margins that actually work. Sometimes boring is the right bet.
Cerebras Systems is targeting an April 2026 Nasdaq listing at a $22-25 billion valuation, aiming to raise roughly $2 billion. Morgan Stanley is the lead underwriter. The company's wafer-scale AI chips are a direct challenge to Nvidia's dominance, and its biggest growth driver is a $10 billion multi-year compute deal with OpenAI -- the largest non-Nvidia AI infrastructure contract ever signed. CFIUS cleared after the company restructured G42's equity stake to non-voting shares.
Scale AI is the wildcard. The data labeling company was last valued at $29 billion following Meta's strategic investment of roughly $14.3 billion for a 49% stake. Founder Alexandr Wang departed to become Meta's Chief AI Officer. Revenue is estimated at $1.5-2 billion with near-100% YoY growth. The Meta deal reduced near-term IPO pressure by providing liquidity, so the timeline is unclear.
The dot-com comparison is unavoidable. And honestly? Parts of it are warranted.
In 1999, over 450 companies went public in US markets, raising a then-record $69.2 billion total. Average first-day gains were 68%. Many companies had no revenue, no business model, no path to profitability.
The 2026 AI wave is structurally different in important ways:
| Dimension | Dot-Com 1999 | AI 2026 |
|---|---|---|
| Number of IPOs | 450+ | Handful of mega-listings |
| Total raised | $69.2B | Potentially $200B+ from just 3-4 companies |
| Revenue | Most had none | OpenAI: $25B, Anthropic: $19B, Databricks: $5.4B |
| Unprofitable tech companies | 36% | ~20% |
| Infrastructure spending | Speculative | $670B from Big Tech alone |
| VC concentration | Broad | 61% of all global VC into AI (2025) |
| P/E ratio (sector) | 45x (Shiller CAPE) | ~22.3x |
The revenue is real. The technology works. The infrastructure is physical, not vapor. These aren't Pets.com.
But there are uncomfortable similarities too.
Sequoia's David Cahn identified a $500 billion annual revenue gap between what AI infrastructure spending implies and what AI companies actually earn. J.P. Morgan projects $5 trillion in AI infrastructure spending through 2030. 61% of all global VC flowed into AI in 2025 -- extreme concentration that creates systemic risk.
And then there's the ROI gap. MIT's study found that 95% of enterprise GenAI pilots achieved zero measurable ROI. An NBER study from February 2026 found 90% of firms reported no impact of AI on workplace productivity. Companies are spending billions on AI and most can't point to a dollar of return.
The smartest people in finance disagree violently on this question. That alone should tell you something.
The bubble camp:
Sam Altman himself said he believes an AI bubble is ongoing. Jeff Bezos called it "kind of an industrial bubble." Paul Kedrosky, a veteran VC, described it as "one of the probably five largest CapEx bubbles in history." Yale's Jeffrey Sonnenfeld warned that the tangle of AI deals among tech giants could be signs of dangerous overinvestment. GMO noted that total AI revenue sits at less than $50 billion against a trillion dollars or more of investment.
40% of CEOs surveyed believe AI hype has driven overinvestment. Goldman Sachs' CEO said publicly: "A lot of capital doesn't deliver returns."
The not-yet camp:
Owen Lamont, a former University of Chicago finance professor now at Acadian Asset Management, uses a "Four Horsemen" framework: overvaluation, bubble beliefs, inflows, and issuance. He says only 3 of 4 conditions exist. The missing piece: equity issuance. Companies did $1 trillion in stock buybacks last year -- the opposite of bubble behavior. The Shiller CAPE is at 40 vs. the dot-com's 45.
Federal Reserve Chair Jerome Powell said the AI sector is "underpinned by substantial realized revenue" and capital expenditure on data centers is "a major engine of broader economic growth."
BlackRock's analysis argues that unlike the dot-com era, today's AI spending is funded by profits and supported by strong balance sheets. US equity funds actually recorded net outflows of ~$45 billion -- the opposite of the retail frenzy that marked 1999.
Here's the thing. A Fortune analysis pointed out that the missing fourth condition -- a wave of IPOs and associated fraud -- could be supplied by the OpenAI IPO itself. The IPO wave might be the event that converts "3 of 4 bubble conditions" into "4 of 4."
No discussion of the AI IPO wave is complete without Masayoshi Son.
SoftBank announced a $100 billion US AI investment alongside Donald Trump, targeting 100,000 new jobs. Son launched "Project Izanagi" -- a $100 billion AI chip venture to challenge Nvidia. He made a $40 billion investment in OpenAI, the single largest check in the $122B round. To fund it, he sold SoftBank's entire $5.8 billion Nvidia stake.
Son is betting on artificial superintelligence arriving within 10 years. SoftBank's stock surged on the announcement, then retreated when investors questioned the strain on the balance sheet. This is the Vision Fund playbook again: massive concentrated bets that either look genius or catastrophic in hindsight. WeWork was the catastrophe. Alibaba was the genius. OpenAI will determine which side of history this round lands on.
If you're considering investing when these companies go public, here's how I'd rank them:
The case: Positive free cash flow. 65% growth. 25x ARR -- expensive but justifiable given the growth rate. Enterprise data platform with genuine switching costs. No consumer product risk. No "we'll be profitable by 2030" hand-waving.
The risk: $134 billion is still a massive valuation for a company with $5.4B in revenue. Competition from Snowflake, AWS, and Google BigQuery is real.
Verdict: If you're going to invest in one AI IPO, this is the one with the least financial risk.
The case: Fastest revenue growth in enterprise software history. 80% enterprise revenue with ~140% net revenue retention. Projects profitability by 2027. Enterprise API share growing while OpenAI's declines. Disciplined spending -- no vanity products.
The risk: $380 billion valuation on $19 billion ARR is 20x revenue. Still burning cash. Infrastructure dependency on Google Cloud TPUs. Revenue concentration among a small number of massive enterprise clients.
Verdict: The strongest margin trajectory of the three. If Anthropic hits its 2027 profitability target, the valuation will look reasonable in retrospect.
The case: Largest AI brand. 910 million users. $25 billion ARR. First-mover advantage in consumer AI.
The risk: Everything else. $14 billion projected loss in 2026. Profitability not until 2030. Declining market share in both consumer and enterprise. Gross margins compressing. Has never disclosed enterprise customer retention rates. At $852 billion, it's valued higher than almost every public company on Earth -- while losing money.
Verdict: The IPO will be a spectacle. The stock will probably pop on day one. But by the time retail investors can buy, VCs and SoftBank will have bought in at valuations 100x lower. The risk-reward for public market investors is the worst of the three.
PitchBook's research is blunt: the market has "rewarded story over substance" -- the highest valuations correlate with the weakest business fundamentals. OpenAI scores weakest on business quality despite having the highest valuation. That's not a statement about the technology. It's a statement about the price.
I think we're watching the formation of the next generation of trillion-dollar companies and the inflation of a bubble. Both things are true simultaneously.
The technology is real. The revenue is real. Anthropic growing from $1B to $19B in 15 months is not a mirage. Databricks generating $5.4B with positive cash flow is not speculation. Even OpenAI's $25B ARR represents genuine value creation.
But the valuations have disconnected from the financials. OpenAI at $852 billion while losing $14 billion per year is not a rational price. It's a bet that the future will be so large that present-day losses don't matter. That bet might pay off. It also might not. And at $852 billion, the margin for error is essentially zero.
Here's what I keep coming back to: 95% of enterprise GenAI pilots achieved zero measurable ROI. Companies are spending billions on AI infrastructure and most can't show a return. The revenue these AI companies are generating is real, but it's heavily concentrated among a few massive customers (Big Tech, financial services, government). The long tail of enterprise adoption -- the thousands of mid-market companies that need to eventually pay for AI to justify these valuations -- hasn't materialized yet.
Analysts identify 2026-2028 as the highest risk window for a significant AI stock correction. I think they're right. The IPO wave itself might be the catalyst. When OpenAI goes public and the S-1 reveals the full extent of its losses, competitive erosion, and margin compression, the narrative could shift fast.
My specific predictions:
Databricks will be the best-performing AI IPO of 2026. Not the flashiest, but the most durable. Positive FCF and enterprise lock-in win over time.
Anthropic will be the most interesting. If they hit the 2027 profitability target, the stock will be a winner. If they miss it, the $380B valuation becomes very hard to defend.
OpenAI will be the most overhyped. The IPO pop will be enormous. The 12-month performance will disappoint. The company needs to grow into a valuation that assumes everything goes right for the next decade.
At least one major AI company will have a down-round or failed IPO attempt by end of 2027. The capital markets aren't infinitely patient, and not every company burning $10B+ per year will find public market buyers.
The dot-com bubble produced Amazon and Google. It also produced Pets.com and Webvan. The AI bubble will produce the next generation of transformative companies. It will also produce spectacular losses for investors who bought the wrong ones at the wrong price.
The trick, as always, is knowing which is which. And right now, I don't think even the companies themselves are sure.