AI Companies Race to IPO as $630 Billion Infrastructure Bet Tests Market’s Risk Appetite

AI companies are flooding public markets at valuations that assume infrastructure returns no enterprise has yet demonstrated, with artificial intelligence names accounting for 92% of the 2026 IPO pipeline value as hyperscalers commit $630 billion to $700 billion in AI infrastructure spending.

SpaceX, following its February merger with xAI, filed for a $75 billion raise at a $1.75 trillion valuation on May 20, targeting a June 11-12 pricing that would mark the largest IPO in history, according to Dealroom. Anthropic confidentially filed its S-1 on June 1 at a $965 billion private valuation with $47 billion in annualized revenue from Claude enterprise sales. OpenAI, valued at $852 billion on $27 billion in annualized revenue, has not yet filed but is expected to list in Q4 2026 or early 2027.

The wave represents peak investor appetite for AI exposure, but the underlying economics remain unproven. None of the four Hyperscalers—Amazon, Alphabet, Meta, Microsoft—have demonstrated positive ROI on their AI infrastructure investments as of Q1 2026 earnings, per Tech Insider. Yet Amazon is committing $200 billion to AI capex in 2026, Alphabet $175-185 billion, Meta $125-145 billion, and Microsoft roughly $120 billion—a collective figure that rivals Sweden’s GDP and nearly doubles 2025’s $380 billion spend.

Cerebras Opens Door, Others Rush Through

Cerebras Systems set the tone on May 14, pricing its IPO at $185 per share before opening above $370—implying a $100 billion fully diluted valuation, more than triple its $23 billion private valuation from February. Institutional demand exceeded available shares by 20 times, forcing underwriters to raise the price range twice and increase share count, Yahoo Finance reported. The stock’s 68% first-day pop signaled that public market investors would pay steep premiums for AI infrastructure exposure regardless of profitability timelines.

Databricks followed on the same day at a $48.8 billion valuation on $510 million in 2025 revenue—a 76% year-over-year increase that justified the premium to data infrastructure investors seeking scaled AI deployment platforms. The company’s S-1, filed April 17, showed enterprise traction across Fortune 500 accounts, giving public investors confidence that AI workloads were generating recurring revenue, not just experimental budgets.

Revenue Growth vs. Capex Reality

The bullish case rests on hyperscaler AI revenue growth that has outpaced baseline cloud expansion. Microsoft Azure AI revenue hit $37 billion annualized in Q1 2026, up 123% year-over-year. Google Cloud revenue surged 63% to $20 billion in the same quarter, with Alphabet citing a $240 billion cloud backlog as evidence of locked-in future demand, per Futurum Research. Amazon Web Services reported AI revenue approaching a $15 billion annual run rate, with Bedrock API calls tripling versus all of 2025.

But capex is scaling faster than revenue. Total AI sector investment reached $202.3 billion in 2025, a 75% increase, while enterprise AI revenue hit just $37 billion—implying a 5.5:1 investment-to-revenue ratio, according to data compiled by Crunchbase. Goldman Sachs warned in April that maintaining acceptable returns would require $1 trillion in annual profit across the AI value chain by 2026, more than double the $450 billion consensus among sell-side analysts. OpenAI alone has committed to $1.4 trillion in capex over eight years despite generating only $13 billion in current revenue.

The ROI Gap That Public Investors Will Inherit

Only 45% of organizations deploying AI can quantify a return on investment, creating a disconnect between infrastructure spending and measurable enterprise outcomes. Amazon’s free cash flow fell to $1.2 billion in Q1 2026 despite $59.3 billion in operating expenses, a dynamic Ferguson Wellman attributed to front-loaded AI capex that won’t generate returns until 2027 at the earliest. The firm noted that capex typically lags ROI realization by 18 months, meaning 2026 spending won’t show up in margin expansion until late 2027 or 2028.

The S&P 500’s capex-to-free-cash-flow ratio currently sits below 1x, well under the 4x peak reached during the dot-com bubble, Fidelity noted. That suggests the AI buildout is occurring on a healthier balance sheet foundation than the late-1990s telecom infrastructure boom. But the comparison offers limited comfort when AI model companies are listing at Valuations that assume enterprise adoption curves steeper than any software category has achieved.

Which Verticals Command Premium Valuations

Inference infrastructure is commanding the steepest multiples. Cerebras, focused on real-time inference workloads, saw its valuation triple in three months. Training infrastructure—the GPUs and clusters used to build foundation models—trades at more modest premiums, as hyperscalers have signaled plans to bring more of that capacity in-house rather than rely on external compute providers.

Model companies with enterprise traction are securing the largest absolute valuations but face the highest scrutiny on retention economics. OpenAI’s enterprise revenue retention rate has never been disclosed, while Anthropic’s $965 billion valuation prices in market share assumptions that The Washington Post noted would require the company to capture a dominant position in enterprise AI across every major vertical.

Data infrastructure platforms—companies like Databricks that sit between cloud compute and enterprise applications—are trading at 30-50x revenue multiples, a premium to traditional SaaS but below pure-play model providers. Investors view these platforms as hedged bets: they benefit from AI workload growth without depending on any single model architecture or training approach to win.

What to Watch

SpaceX’s June 11-12 pricing will test whether trillion-dollar valuations remain achievable after Cerebras’s first-day pop fades. If the stock trades flat or down in the weeks following Cerebras’s debut, underwriters may pressure SpaceX to reduce its valuation target or raise size. Anthropic’s S-1, once publicly filed, will reveal enterprise gross margins and customer concentration—metrics that will determine whether model companies can achieve software-like economics or remain capital-intensive services businesses.

Monitor Q2 2026 earnings from hyperscalers for any capex guidance revisions. If infrastructure spending growth decelerates before AI revenue growth accelerates, the investment-to-revenue gap will widen further, pressuring public market valuations across the sector. Amazon CEO Andy Jassy’s May comment—“We’re not going to be conservative in how we play this”—suggests hyperscalers remain committed to the buildout regardless of near-term ROI visibility, but AWS free cash flow trends will determine how long that commitment can be sustained without margin compression triggering investor concern.