Anthropic’s potential IPO may be one of the biggest technology listings in years, but a Wall Street veteran believes investors should look beyond the artificial intelligence company’s headline revenue growth. The bigger issue, he argues, could be revenue quality, and a structural mismatch between Anthropic’s largely fixed costs and its usage-driven revenue.

Dr. Chan Ahn, founder and CEO of Tessera PE and a former Goldman Sachs and JPMorgan executive, told Benzinga that Anthropic’s business could face a problem familiar to investors who watched the telecommunications boom of the late 1990s: companies can lock themselves into enormous long-term capacity commitments just as the economics of that capacity begin to deteriorate.

That could make Anthropic’s reported growth look considerably stronger than the underlying economics of its business.

What is the biggest risk facing Anthropic before its IPO?

According to Ahn, the key question is not simply how much revenue Anthropic generates. It is how durable that revenue is.

The company’s cost structure reportedly includes large, long-term commitments for computing capacity, while much of its revenue comes from customers paying according to their usage of its AI models.

That creates a potentially important mismatch:

• Costs: Large portions are fixed or committed for years.
• Revenue: A significant portion is tied to usage and can change rapidly.
• Pricing: AI usage can be repriced as competition increases and model costs decline.
• Demand: Customers can potentially reduce usage much faster than Anthropic can reduce its infrastructure commitments.

For a private company valued on future growth, that risk can remain difficult to see. Public investors, however, typically place much greater emphasis on recurring revenue, margins, cash flow, and the quality of contracted business.

Why the distinction matters

A customer signing a multi-year enterprise contract provides greater visibility than a customer simply using an AI model on demand.

If Anthropic has a large base of committed enterprise revenue, its future cash flow could be more predictable. If a substantial portion of its revenue instead comes from usage that customers can quickly reduce, the company could face greater volatility.

That is why Ahn believes the split between committed enterprise agreements and pure on-demand usage could become one of the most important divs in Anthropic’s eventual S-1 filing.

Consider adding an infographic here showing fixed infrastructure commitments versus usage-based revenue, illustrating how a fall in customer demand could affect margins.

Could Anthropic’s computing commitments become a problem?

Ahn highlighted what he described as a roughly $1.25 billion-a-month commitment extending through May 2029.

That would translate into approximately $15 billion annually. The reported commitment is particularly notable because it is associated with SpaceX following its acquisition of xAI, putting Anthropic in the unusual position of making a substantial infrastructure commitment involving a major player in the AI industry.

The important issue is not simply the size of the contract.

It is the fact that Anthropic could remain obligated to pay for capacity even if the amount customers are willing to pay for access to AI models changes significantly.

That creates what financial analysts sometimes call a stranded-cost risk: a company continues paying for capacity that no longer produces the expected economic return.

What does the telecom boom of 1999-2002 have to do with Anthropic?

Ahn compares Anthropic’s situation with the long-haul telecommunications buildout around the turn of the century.

During the late 1990s and early 2000s, telecommunications companies invested heavily in fibre-optic networks based on expectations that internet traffic would grow rapidly.

The problem was that capacity expanded faster than demand in many parts of the market.

As additional capacity came online, bandwidth prices fell sharply. Companies that had committed to expensive, long-term capacity contracts could therefore find themselves paying for infrastructure whose market value had fallen substantially.

A similar dynamic could theoretically emerge in AI.

If computing capacity becomes abundant while the cost of inference falls rapidly, Anthropic could be locked into expensive commitments negotiated when AI demand and pricing expectations were considerably higher.

The analogy is not proof that Anthropic will experience the same outcome. The AI market differs significantly from telecommunications, particularly because AI demand is still expanding rapidly and computing requirements continue to rise.

But the comparison highlights a risk investors will need to evaluate: rapid technological progress can increase demand while simultaneously destroying the economics of older capacity.

Could regulation suddenly disrupt Anthropic’s revenue?

Ahn also points to regulatory risk as another reason investors should scrutinize Anthropic’s usage-based revenue model.

The argument is straightforward. If regulation, export restrictions or government action prevents customers from accessing particular AI models, usage-based revenue can fall almost immediately.

Anthropic’s infrastructure commitments, however, may not decline at the same speed.

That creates an asymmetry.

A customer can stop or reduce usage quickly, while a multi-year infrastructure agreement may remain in place.

Ahn summarized the problem by contrasting a metered service with a fixed obligation: customer demand can change rapidly, while contractual costs may remain unchanged.

This is particularly relevant for an AI company operating at the intersection of commercial technology, national security and increasingly restrictive export controls.

Revenue growth is often one of the most attractive numbers in an emerging technology company.

But public investors eventually ask harder questions:

1. How much revenue is contractually committed?
2. How much can disappear if customers reduce usage?
3. What does it cost to generate each dollar of revenue?
4. How quickly are inference costs falling?
5. How much computing capacity has the company committed to?
6. What happens to margins if AI prices decline?
7. How much cash must be invested to support future growth?

For Anthropic, those questions could matter more than a headline annualized revenue number.

An AI company can grow extremely quickly while still consuming enormous amounts of capital. If revenue grows at 100% but the cost of serving that revenue grows almost as quickly, investors may eventually question whether the growth creates enough economic value.

Is Anthropic being valued like software company or a research laboratory?

Ahn’s broader argument concerns the unusual financial structure of frontier AI companies.

Companies such as Anthropic are simultaneously:

• Building commercial software products;
• Operating enormous computing infrastructure;
• Conducting cutting-edge AI research;
• Competing against some of the world’s largest technology companies;
• Spending heavily before the long-term economics of frontier AI are fully established.

That creates a difficult valuation problem.

Traditional software companies can often scale revenue with comparatively low incremental costs. Frontier AI companies generally require significant computing resources to train and operate increasingly capable models.

The result is that Anthropic may resemble a software company from a revenue perspective while carrying some of the financial characteristics of a capital-intensive research operation.

Ahn argues that an IPO does not eliminate that underlying scientific and technological risk. Instead, public markets inherit the risk.

That distinction could become increasingly important if investors begin valuing AI companies on conventional software metrics without adequately accounting for their infrastructure requirements.

What does Anthropic’s $2 trillion valuation imply?

Anthropic’s reported valuation has already reached extraordinary levels, making the company’s potential IPO particularly consequential.

Ahn argues that a valuation around $2 trillion would require exceptional revenue growth and significant margin expansion over many years.

His estimates suggest that Anthropic could need hundreds of billions of dollars in annual revenue by 2036 to justify such a valuation under different discount-rate assumptions.

Those divs are valuation scenarios, not forecasts of Anthropic’s actual future revenue. Investors should therefore treat them as a way of illustrating the scale of expectations embedded in the valuation rather than as a prediction.

The company’s reported annualized revenue run rate has also climbed dramatically, with Benzinga citing a div of approximately $65 billion.

But revenue alone does not determine whether the valuation is sustainable.

The more important question may be whether Anthropic can convert that growth into durable free cash flow while keeping computing and research costs under control.

Could Anthropic’s IPO face the “public market reality check”?

Private-market valuations can reflect expectations about future technological breakthroughs, market expansion and strategic importance.

Public markets are less forgiving.

Once listed, Anthropic would have to provide investors with regular financial disclosures and face continuous scrutiny over:

• Revenue growth;
• Gross margins;
• Operating expenses;
• Capital expenditures;
• Cash burn;
• Customer concentration;
• Contractual commitments;
• Model pricing;
• Computing costs;
• Free cash flow.

That could produce a very different valuation conversation from the one that takes place in private funding rounds.

A company can be strategically important and technologically impressive without necessarily generating the financial returns required to justify an enormous public-market valuation.

Why SpaceX and xAI matter to the Anthropic story

The reported infrastructure commitment involving SpaceX adds another layer to the story.

SpaceX’s acquisition of xAI combines two major technology businesses with ambitions across space, artificial intelligence and computing.

For Anthropic, however, the relationship highlights an important competitive reality: AI companies are increasingly dependent on enormous pools of computing power, and access to that infrastructure can become a strategic asset.

That makes infrastructure contracts more than simple supplier agreements.

They can influence a company’s margins, capacity, competitive position and financial risk for years.

Is Anthropic’s $30 trillion opportunity realistic?

Another report cited by Benzinga says Anthropic is preparing to target a potential revenue opportunity exceeding $30 trillion, larger than an estimated $28.5 trillion opportunity associated with SpaceX.

These divs should be treated cautiously.

A “revenue opportunity” or total addressable market is fundamentally different from actual revenue. A massive theoretical market does not mean a company will capture anything close to that amount.

The more useful question for investors is how much of that opportunity Anthropic can realistically monetize.

Its strongest argument may not be model leadership alone. The company will need to establish itself deeply inside enterprise workflows, where customers are more likely to build AI into recurring business processes.

That could make enterprise adoption a more durable competitive advantage than simply having the most capable model at a particular moment.

What should investors watch in Anthropic’s S-1?

If Anthropic files for an IPO, investors will have an opportunity to test many of Ahn’s concerns against the company’s actual financial disclosures.

The most important areas to examine will include:

1. Contracted versus usage-based revenue

This may provide the clearest indication of revenue durability.

2. Customer concentration

A small number of large customers could create significant revenue risk if one reduces its usage or switches providers.

3. Computing commitments

Investors should examine the duration, pricing and minimum-payment obligations associated with infrastructure contracts.

4. Gross margins

Rapid revenue growth means less if the cost of delivering AI services remains extremely high.

5. Free cash flow

This will show whether Anthropic’s growth is beginning to generate cash or remains dependent on external capital.

6. Capital expenditures and future commitments

AI companies can require enormous investment in computing infrastructure. Existing obligations may not capture the full amount required to support future growth.

7. Pricing trends

If AI inference prices continue falling, Anthropic will need substantial volume growth to offset lower revenue per unit of usage.

What is really at stake for Anthropic?

Anthropic’s potential IPO is not simply a test of whether investors believe in Claude or the future of AI.

It is also a test of whether the economics of frontier AI can support the valuations now being placed on the sector.

The bullish case is straightforward: AI adoption could expand dramatically, enterprise customers could embed models into critical workflows, computing efficiency could improve, and Anthropic could turn its rapidly expanding revenue base into substantial profits.

The bear case is equally important: AI prices could fall faster than demand grows, customers could shift between competing models, infrastructure commitments could become expensive burdens, and research spending could remain enormous.

That is why revenue quality may ultimately matter more than revenue growth.

Anthropic could enter the public markets with one of the fastest-growing revenue profiles ever seen from a technology company. But investors will still have to answer a much harder question:

How much of that revenue is durable, profitable, and capable of supporting the enormous fixed commitments required to remain at the frontier of AI?

The answer could determine whether Anthropic’s IPO becomes a landmark public-market debut—or an early warning about the financial risks of the AI boom.