The AI funding window is closing: slowing revenue, IPO fails, and a bankrupt OpenAI

For two years, Silicon Valley and global equity markets agreed that artificial intelligence (AI) is an uncapped frontier where aggressive capital deployment and unchecked spending would automatically yield exponential returns. Or, perhaps, equity markets believed the spin coming out of Silicon Valley.

Whichever scenario is true, reported operating results over the past week have confirmed the shift in sentiment toward AI.

Slowing revenue, public market pullbacks, and fluid enterprise strategies suggest the market’s appetite for unconstrained AI spending is finally hitting a ceiling. Investors have well and truly moved past the honeymoon phase and are demanding proof of a sustainable economic model.

Slowing growth

The first major signal came directly from Anthropic, one of frontier labs in the race to produce the smartest AI models.

After a closed-door pre-Initial Public Offering (IPO) investor meeting in mid-October 2026, attendees revealed the company reported US$16.8 billion in Q3 revenue (up 45-50 per cent quarter-over-quarter) and it achieved adjusted profitability. The growth rate however marked a sequential deceleration from Q2, when revenue surged by more than 140 per cent.

According to attendees, a notable portion of the company’s previous growth rates was driven by ‘model distillation’ – a practice where rival AI labs and companies created thousands of accounts to systematically prompt Claude and use its answers to train their own competing models. Anthropic said it actively implemented restrictions, cutting back on distillation revenue, to stop rivals from scraping its intelligence and copying its proprietary AI secrets, sacrificing short-term revenue growth for long-term IP protection.

For investors, the issue isn’t that individual consumers are reluctant to pay subscription fees for AI assistants when free tiers are available, but that AI-sector valuations built on the assumption of massive enterprise growth rates are unjustified.

Corporate buyers – recently expected to spend generously on AI tokens indefinitely – are beginning to scrutinise their operational budgets.

Polymarket, tokenmaxxing and open source

For many investors, Anthropic’s latest financials also point to a shift in enterprise customer behaviour. During the initial wave of AI model deployment, corporates engaged and experimented heavily in ‘tokenmaxxing’ – the practice of passing vast amounts of context, raw data, and brute-force iterations through expensive frontier proprietary models to complete complex workflows.

Now, companies such as Uber and Skywork.ai are actively re-engineering for cost efficiency. Enterprises are scaling back raw context ingestion, tightening rate limits, and increasingly turning to open-source models.

Open-source alternatives are now becoming highly capable and allow organisations to run specialised tasks on local compute or cheaper cloud instances at a fraction of the cost of commercial Application Programming Interfaces (APIs).

Many fear this transition will directly affect the revenue trajectories of top-tier model providers.

Indeed, on prediction markets like Polymarket, probabilities around Anthropic’s IPO have noticeably repriced. Retail speculators once viewed a valuation above US$2 trillion as a baseline for the company’s float, but that is no longer consensus.

Figure 1. Bets on Anthropic’s IPO Valuation

Infrastructure Volatility and the IPO Market

Anthropic isn’t the only AI-related company finding the path to market more challenging.

South Korean neocloud firm Elice Group made its public trading debut on the KOSDAQ market this week. Despite strong initial excitement – retail investors oversubscribed the IPO more than 400 times – the stock plummeted 32 per cent on its first day of trading.

Figure 2. Elice Inc. debut on KOSDAQ

Elice builds modular data centres, and rents specialised Graphic Processing Unit (GPU) capacity. This was once an exciting place to invest – the frontline, levered to the AI boom. But Elice Group’s share plunge this week shows sentiment has pivoted, and investors are growing more cautious about the infrastructure buildout, questioning how quickly the hundreds of billions spent on GPUs, liquid cooling, and data centres will translate into cash flow.

Could OPENAI be bankrupted?

Beyond IPO’s, Tokemaxxing and enterprise cost optimisation, a debate has also begun taking root among financial journalists, investors, pundits, short-sellers, and tech analysts about the long-term solvency of frontier model providers like OpenAI.

Tech columnist Ed Zitron points to OpenAI’s cash burn – US$21.6 billion in the first half of 2026 and an operating loss of US$12.3 billion in 2Q26 – arguing the company’s current financial trajectory resembles a “capital sinkhole”.

If enterprises keep migrating to cheaper open-source alternatives and closed-model revenues compress, ongoing multi-billion-dollar private funding rounds will eventually become unsustainable.

While most mainstream institutions don’t expect a sudden, catastrophic liquidation – predicting instead that key financial partners like Microsoft would intervene or restructure equity – the fact that OpenAI’s solvency is being debated marks an inflection point in the AI bubble’s narrative.

If this argument gains traction across public and venture capital markets, it will further shift sentiment away from supporting the broader AI boom, and result in tightening liquidity for the very companies being debated.

Investor concern over the structural economics of model providers therefore risks triggering a wider pullback in capital deployment, cooling speculative valuations across hardware and neocloud sectors (we’re already seeing this), and forcing the industry into a period of fiscal conservatism that would only help burst the bubble.

Not over yet

Importantly, the shift in sentiment has so far been modest. It does not reflect an outright collapse of the AI industry, though that may still come. Instead, the shift marks a move away from speculative land-grabbing and toward disciplined economic fundamentals. Whether it pauses here will depend on whether the AI buildout can indeed deliver economic returns.

We’re at a reassessment point where investors are setting what they believe are realistic limits on the valuation multiples AI companies can command, and they want clear paths to long-term profitability rather than relying on more dubious revenue expansion alone.

If AI frontier lab customers learn to achieve equal performance with smaller, distilled models rather than overspending on massive context windows, retail consumers remain price-sensitive about monthly subscriptions, and IPO’s continue to fail or be deferred, the current market adjustment may not be over.

And that would mean the era of frictionless funding is over. The winners of the next phase will be the companies that prioritise cost efficiency, clear utility, and sustainable capital management, but the entire sector may need to collapse first.