Anthropic is entering the final stretch of one of the most consequential initial public offerings (IPOs) in technology history, setting the stage for a high-stakes market test of whether massive artificial intelligence (AI) infrastructure spending can translate into durable profits.

The San Francisco-based startup is slated to price its shares Wednesday night, ahead of a Nasdaq debut under the ticker symbol ANT on Thursday.

The IPO comprises 825 million Class A shares marketed at $70 to $78 apiece, with the goal to raise up to $64.4 billion. That pricing pegs Anthropic at a fully diluted valuation of roughly $1.3 trillion to $1.4 trillion, a notable drop from the $2 trillion figure rumored in September.

Early demand appears robust: Bloomberg reported last week that the book is already roughly six times oversubscribed. Alongside the public float — which reserves roughly 20% for retail investors subject to a 15-day anti-flipping window — Anthropic has secured an $81 billion concurrent private placement from seven corporate backers, including NVIDIA Corp.

According to analysts at Wedbush and Jefferies, the valuation reset is a calculated bid to make the massive offering digestible, ensure a strong post-listing pop, and avoid a weak debut that could rattle broader AI equities.

“Anthropic’s IPO is the first time public investors get to see an AI lab’s own numbers, and the numbers are real: $17 billion of revenue in the third quarter alone. What stands out is that the growth comes from volume, not price,” said Michael Semenov, chief investment officer at DPRG Investment Management.

“Revenue earned per dollar of compute rose from $0.40 to $2.08 in six quarters, even as the price per token keeps falling,” Semenov said. “The question for investors is no longer whether the business works, but what they pay for it: the compute contracts are owed whatever revenue does, and we expect 2027 to be a loss-making year as that capacity comes on ahead of revenue.”

Still, the roadshow has exposed a sharp divide on Wall Street over how to value a business growing at breakneck speed while carrying unprecedented capital commitments.

Anthropic’s top-line momentum is undeniable. Through the first nine months of 2026, revenue surged more than tenfold year-over-year to $33.2 billion, driven heavily by consumption-based API usage, which accounts for 79% of sales, and a roster of nearly 12,000 enterprise customers spending over $100,000 annually.

“Given the data we see in enterprises we are extremely bullish on the Anthropic IPO,” said Russ Fradin, co-founder and CEO of AI measurement platform Larridin. “We are still in the very early innings of AI adoption in the enterprise and it is clear that customers who use Anthropic products use them in more sophisticated ways over time. We expect them to have amazing success.”

However, financing and accounting adjustments pushed the company’s nine-month GAAP net loss to $104 billion, despite briefly turning an operating profit in the second quarter.

The central anxiety among institutional buyers is Anthropic’s $895.4 billion in multi-year compute and infrastructure commitments, which include massive contracts with Alphabet Inc.’s Google ($114.5 billion), Amazon.com Inc. ($110 billion), Microsoft Corp. ($45 billion), and Elon Musk’s SpaceX (up to $120.5 billion).

During an investor roadshow last week, Anthropic CEO Dario Amodei acknowledged that forecasting compute demand remains a “huge uncertainty,” though he pointed to strong unit-compute gross margins and a fortress balance sheet — boasting over $95 billion in cash and $36 billion in short-term investments prior to IPO proceeds — as critical buffers.

Political and physical bottlenecks also loom. Amended S-1 filings warn that power grid constraints, local opposition to data centers, and post-midterm regulatory shifts could delay the very capacity Anthropic is contracted to fund.

Equity researchers and market practitioners remain split on whether the IPO price represents a bargain or a trap.

Morningstar analyst Malik Ahmed Khan argues the market’s caution around capital expenditures has left the stock undervalued, assigning it a $92 per share ($1.7 trillion) fair value estimate based on projected 47% annual revenue growth over five years and a narrow economic moat built on frontier research and multi-chip efficiency.

Yet Morningstar’s Very High Uncertainty rating spans from a $188 bull case to a $25 bear case should open-weight models commoditize the sector.

For some seasoned operators, the unit economics counsel restraint.

“They keep about 85 cents of every dollar after serving the customer. Then 58 cents goes back into research, and they can’t skip it,” said James Deller, an investor, operator, and technology entrepreneur whose team relies heavily on Anthropic’s Claude models.

Noting that more than 80% of last quarter’s revenue came from models less than six months old, Deller views continuous research as an unavoidable cost of sale. “At roughly 40 times sales, investors are paying for margins the company hasn’t yet demonstrated,” he added, comparing the setup to Facebook’s post-IPO stumble. “I like the company. I just think the price has got ahead of it.”

Pete Bernard, CEO of the EDGE AI Foundation, said the offering unfolds against two major trends: mounting risks from AI consolidation among a few dominant players, and rapid ecosystem diversification driven by open models and ten-fold edge AI performance gains every six months.

Questioning how Anthropic will navigate these dynamics alongside growing pushback against massive data center capital expenditures, Bernard added that the edge AI sector is experiencing tremendous growth and collaboration—and welcomed Anthropic’s future participation.

While the filing highlights the massive operational costs of running a frontier AI firm, Flexera Chief Financial Officer Chris Andersen noted that most enterprises still lack basic visibility into what it actually costs to use the technology. He warned that declining token prices driven by market competition will not curb corporate spending. Instead, total bills continue to climb as businesses deploy more autonomous agents, expand automated workflows, and upgrade employees to premium models.

Furthermore, AI providers face mounting pressure to convert heavy compute commitments into sustainable revenue, which will likely trigger unpredictable shifts in pricing models. Andersen urged fellow CFOs to audit their AI investments immediately — tracking every dollar spent, assigning clear ownership, and measuring returns — before sudden vendor price changes force their hand.

Even skeptics of the valuation, however, welcome the transparency public markets will bring.

As Deller noted, quarterly audited financials will offer enterprise customers vital visibility into a core vendor — even as risk disclosures reminding buyers of potential rate-limit changes reinforce the need to keep AI workflows portable.

This week, public investors will finally render their verdict.