Claude’s annualized revenue pace has risen more than sevenfold since the end of 2025. The harder question is whether that growth is durable enough to support a near-$1 trillion valuation — and Anthropic’s $190–$200 billion revenue target for 2028.
Anthropic’s revenue run rate has risen at a pace that would look extraordinary even by the standards of the current AI boom. The Claude maker’s annual revenue run rate exceeded $65 billion by the end of July 2026, according to Reuters, up from $47 billion in May and approximately $9 billion at the end of 2025. That represents a more than sevenfold increase in roughly seven months.
Investors are clearly paying attention. Anthropic raised $65 billion in May at a $965 billion post-money valuation, then confidentially filed for a U.S. initial public offering on June 1. Meanwhile, people familiar with the company’s financial projections told Reuters that Anthropic is forecasting roughly $190 billion to $200 billion of revenue in 2028.
Those numbers make the growth story difficult to ignore. But they also make the underlying economics more important. Both things can be true at once: Anthropic’s commercial traction is extraordinary, and a $65 billion revenue run rate is not the same thing as $65 billion of durable, audited annual revenue.
The real question is therefore not whether Anthropic is growing. It clearly is. The question is whether the quality, durability and economics of that growth can support the valuation assumptions now being constructed around one of the most closely watched technology IPOs.
The $65 Billion Number Needs Context
The headline figure is a revenue run rate above $65 billion as of the end of July 2026. That was up approximately 38% from the $47 billion annualized pace Anthropic disclosed in May, in just two months, and more than seven times the roughly $9 billion run rate reported at the end of 2025.
Reuters reported that Anthropic shared the updated figure with investors as part of ongoing financial communications, while Bloomberg first reported the latest number publicly.
Run rate, however, deserves careful definition because it is not the same as annual revenue. A revenue run rate takes a company’s current sales pace and extrapolates it over a full year. In a rapidly growing business, that can provide a useful picture of direction, but it can also make growth appear smoother and more permanent than it ultimately proves to be.
That distinction matters especially in AI. A conventional enterprise software company may sell customers multiyear contracts with relatively predictable recurring subscription revenue. Generative AI economics can be more variable. Usage can rise sharply when enterprises deploy new AI agents or coding systems, but consumption can also respond to model prices, token costs, product changes, competing models and shifts in customer workloads.
That makes the $65 billion figure simultaneously remarkable and incomplete.
Claude Code Is Becoming a Major Commercial Engine
A major contributor to Anthropic’s momentum appears to be Claude’s position in software development and enterprise AI. Claude Code has emerged as one of the company’s most strategically important products, helping Anthropic gain traction among developers and businesses that are embedding AI directly into engineering workflows.
That matters because coding is fundamentally different from casual chatbot usage. A consumer can switch between Claude, ChatGPT and Gemini in seconds. An enterprise that has built processes, internal tools, evaluation systems and developer workflows around one model faces considerably more friction.
That does not make customers captive, but it does create a potential layer of switching cost that consumer AI products do not possess. The stronger Claude becomes inside software-development workflows, the more Anthropic begins to resemble an enterprise infrastructure provider rather than simply another chatbot company.
That distinction could matter enormously for its IPO valuation. It is also why Anthropic’s recent infrastructure moves deserve to be viewed alongside its revenue growth. Blockgeni recently examined Anthropic’s potential $6 billion acquisition of Decart AI, a deal that could give the company greater control over inference efficiency and the cost of running Claude.
The connection is straightforward: revenue growth determines how valuable Claude becomes, while inference economics determine how much of that revenue Anthropic can eventually keep.
The AI Race Is Moving Beyond Model Benchmarks
Anthropic’s trajectory reflects a broader change in generative AI. The competitive conversation is gradually moving away from a single question — which company has the smartest model? — toward much harder questions about monetization, customer retention, infrastructure economics and capital efficiency.
The next phase of the AI race will increasingly be determined by which companies can monetize their models, retain enterprise customers, run those models cheaply enough to generate attractive margins and finance the infrastructure required to support rapidly expanding demand.
Those questions become much more important once frontier AI companies enter public markets. Private investors can underwrite aggressive long-term forecasts. Public markets eventually demand audited revenue, gross margins, operating cash flow, customer concentration, capital expenditure, retention and a clearer path to profitability.
Anthropic has now confidentially filed for an IPO. OpenAI subsequently filed confidentially as well. That means the world’s two most prominent independent frontier AI companies are moving toward public-market scrutiny at almost the same time.
The result could fundamentally change how the AI race is measured. Model benchmarks will still matter, but earnings reports may soon matter more.
The $190–$200 Billion Question
Anthropic’s reported 2028 revenue forecast is therefore central to the story. Reuters reported that investors and bankers are evaluating a projection of roughly $190 billion to $200 billion of revenue in 2028 as they consider what Anthropic could be worth when it reaches public markets.
At first glance, that target looks almost impossibly large. But the mathematics is more nuanced than the headline suggests. From a current $65 billion annualized pace, reaching approximately $195 billion would require the business to roughly triple.
That is still extraordinarily demanding for a company already operating at this scale, but Anthropic would not need to repeat its sevenfold increase from the past seven months.
The real challenge is sustaining substantial growth while the industry around Anthropic becomes more competitive. Model prices are falling, inference is becoming cheaper, open-weight alternatives are improving, and Google, Meta, OpenAI and other competitors are investing enormous sums.
Enterprise buyers are also becoming more sophisticated about comparing models based on cost, performance, latency and reliability. Anthropic therefore needs to roughly triple its business while simultaneously defending pricing and improving the economics underneath that revenue.
That is a much more useful way to frame the $200 billion question.
A $965 Billion Valuation Raises the Standard
Anthropic’s May financing makes the analysis even more important. The company raised $65 billion in its Series H round at a $965 billion post-money valuation.
At a $65 billion annualized revenue run rate, that implies a valuation of roughly 15 times current run-rate revenue. That multiple is not automatically irrational for an exceptionally fast-growing technology business, but it embeds enormous assumptions.
Investors are effectively betting that Anthropic can maintain rapid enterprise adoption, preserve substantial pricing power, convert growing usage into durable customer relationships, control infrastructure costs and eventually produce margins worthy of one of the most valuable technology companies in the world.
The last point may prove the hardest.
Generative AI is extraordinarily capital intensive. Every additional enterprise customer creates revenue, but it also creates demand for inference. Unlike conventional software, serving an additional AI request involves meaningful computing costs.
That means revenue growth cannot be evaluated separately from infrastructure efficiency. Blockgeni’s analysis of Nvidia’s effort to mobilize more than $500 billion of third-party capital for AI compute illustrates just how dramatically the financing model underneath AI is already changing.
Anthropic’s revenue story sits on top of that same physical infrastructure.
Anthropic’s Infrastructure Commitments Are Growing Just as Fast
This is why Anthropic’s infrastructure commitments matter so much. Amazon and Anthropic dramatically expanded their relationship in April. Amazon agreed to invest $5 billion immediately and potentially another $20 billion over time, in addition to the $8 billion it had previously invested.
At the same time, Anthropic committed to spend more than $100 billion on Amazon’s cloud infrastructure over the next decade, securing access to as much as 5 gigawatts of computing capacity.
That relationship creates both strength and dependency. AWS provides Anthropic with enormous computing capacity and distribution, while Anthropic provides Amazon with one of the fastest-growing customers in the AI economy and a major user of Amazon’s custom Trainium processors.
But it means that investors analysing Anthropic’s revenue should also ask a more difficult question: how much does it cost Anthropic to generate each additional dollar of that revenue?
That question becomes even more important as usage grows.
Anthropic is simultaneously securing large-scale physical capacity elsewhere. Blockgeni recently examined the Riot Platforms–Anthropic $9.1 billion infrastructure agreement, showing how the Claude maker is locking in enormous amounts of data-centre and power capacity to support future expansion.
It is also considering deeper control over compute optimization through the reported Decart negotiations. Taken together, these moves suggest Anthropic understands that the next competitive advantage in frontier AI may not simply be model intelligence.
It may be cost per useful unit of intelligence.
Consumption Revenue Can Grow Quickly — and Reprice Quickly
There is another tension inside the $65 billion number. One increasingly important contributor to Anthropic’s business appears to be developer and enterprise AI consumption.
Usage-based revenue can be extremely attractive. When customers use more Claude, Anthropic earns more. When enterprises deploy agents across thousands of employees or millions of automated operations, consumption can expand dramatically without requiring a traditional seat-by-seat software rollout.
But usage-based businesses also expose companies more directly to pricing competition.
Suppose another frontier model reaches comparable coding performance at meaningfully lower cost. Enterprise procurement teams will notice. That does not mean customers immediately abandon Claude. Changing production AI systems can involve substantial integration work, testing, security validation and employee retraining.
But it can still influence contract negotiations and the pricing of future workloads.
That is why Anthropic’s revenue quality matters almost as much as its revenue growth. The important question is not simply how many dollars Claude is generating today. It is how defensible those dollars remain when models become cheaper and capability gaps narrow.
Hyperscaler Dependency Cuts Both Ways
Amazon is not Anthropic’s only infrastructure partner. The company has also expanded its relationship with Google and Broadcom for multiple gigawatts of next-generation TPU capacity.
Maintaining relationships across several computing ecosystems rather than betting everything on one semiconductor architecture makes strategic sense. But the broader hyperscaler relationship remains complicated.
AWS, Google Cloud and Microsoft Azure are simultaneously distribution channels, infrastructure providers, strategic investors or partners, and competitors building their own AI systems.
That creates a form of dependency that traditional software companies rarely face at this scale.
Anthropic needs hyperscalers to distribute and run Claude. Those same hyperscalers have incentives to strengthen their own models. Google can push Gemini. Amazon continues to develop its own AI models and custom silicon. Microsoft remains deeply connected to OpenAI while expanding its broader model ecosystem.
Anthropic therefore needs partners that could eventually become more aggressive competitors. Public-market investors will almost certainly scrutinize that relationship.
The Strongest Counterargument
There is a strong argument against being overly skeptical about the run-rate figure.
Enterprise customers do not necessarily treat frontier models as interchangeable commodities. Once an organization integrates Claude into software-development workflows, internal applications, security processes, automation systems or agent architectures, switching providers can become operationally expensive.
Applications have to be retested, prompts and workflows may behave differently, model outputs need to be revalidated, employees may need retraining, security teams may need to approve a new provider, and production systems may have to be rebuilt around different APIs and operational characteristics.
That friction is real.
The broader strategic stakes of model dependency are now becoming important enough that governments themselves are examining them. As Blockgeni recently reported, the UK government is studying the economic consequences of losing access to frontier AI models.
That is a striking signal. If access to a frontier model can become important enough to constitute a national economic dependency, switching costs at the enterprise level are unlikely to be trivial.
That strengthens the bullish argument for Anthropic, but it does not eliminate pricing risk. A product can be sticky while still facing pressure on the price charged for new usage.
That distinction may become central to Anthropic’s eventual public-market valuation.
Safety Is Part of the Revenue Story Too
Anthropic has built much of its brand around responsible and safety-focused AI development. That positioning can strengthen enterprise trust, but it also creates a higher reputational standard.
The more companies delegate meaningful work to Claude agents, the more model reliability becomes a commercial issue rather than merely a research question.
Blockgeni recently examined Anthropic’s own research showing AI agents can obscure aspects of their reasoning or behavior.
The importance for investors is straightforward. An enterprise AI company valued near $1 trillion is no longer evaluated only on whether its models are technically impressive.
Reliability failures can affect customer retention, insurance requirements, enterprise procurement, regulatory scrutiny and ultimately revenue.
Safety therefore sits directly inside the valuation story.
The Risks Behind the Growth Story
One of the biggest risks is open-weight model parity. If open models approach Claude’s coding and agent capabilities while remaining significantly cheaper to operate, enterprise buyers could gain substantially more negotiating power.
Pricing compression presents another risk. Frontier-model capability continues improving while inference costs fall. That is positive for adoption, but it may make today’s pricing difficult to maintain indefinitely.
Hyperscaler vertical integration could create further pressure. Amazon, Google and Microsoft all have strategic incentives to promote their own AI infrastructure and models, meaning Anthropic must remain strategically valuable to partners that also compete with it.
Infrastructure costs also remain crucial. Anthropic’s growth requires enormous spending on chips, data centres and electricity. Strong revenue growth does not automatically translate into attractive free cash flow.
IPO conditions could further complicate the valuation story. A deterioration in public equity markets or a major disappointment elsewhere in the AI sector could reduce the multiple investors are willing to apply.
Regulatory and safety exposure also matter because Anthropic’s safety-focused reputation makes significant failures particularly consequential.
Finally, run-rate normalization remains possible. A revenue run rate based on unusually strong recent consumption could moderate even if the underlying business remains healthy.
None of these risks implies Anthropic’s growth is artificial. They determine how much investors should be willing to pay for it.
What Investors Should Watch Before the IPO
The first number to watch is actual reported revenue. Once Anthropic’s public prospectus becomes available, investors should be able to compare annualized run-rate figures with historical audited revenue. That comparison will immediately reveal how closely the headline growth trajectory reflects recognized sales.
The second is gross margin. A $65 billion AI business with software-like margins deserves one valuation. A $65 billion business requiring enormous ongoing compute expenditure deserves another.
Customer concentration will also matter. If a relatively small group of large enterprise or API customers accounts for a disproportionate amount of growth, the run-rate figure becomes more sensitive to individual contract decisions.
Retention and expansion should be equally important. How much more are existing Claude customers spending over time? That number may ultimately be more valuable than the headline count of new users.
Finally, investors will need to examine capital expenditure and infrastructure commitments. Anthropic’s infrastructure expansion is rapidly becoming one of the largest strategic components of the company, and markets will want to understand how those commitments translate into future margins.
Where This Ends Up
The most plausible outcome is not that Anthropic’s extraordinary revenue growth suddenly disappears. The company has demonstrated genuine commercial traction, particularly around coding and enterprise AI, and its infrastructure expansion suggests management expects demand to continue rising substantially.
But the next phase of the story will be different from the last.
Private investors rewarded acceleration.
Public markets will ask what that acceleration costs.
A potential Anthropic listing later in 2026, subject to SEC review and market conditions, could therefore become one of the first major tests of how investors value a frontier AI company once they can see more of the actual economics underneath the models.
If Anthropic can demonstrate that its $65 billion run rate is backed by durable enterprise customers, strong retention and improving margins, the $190–$200 billion 2028 forecast becomes considerably easier to defend.
If the eventual prospectus reveals that extraordinary growth is accompanied by equally extraordinary compute expenditure, customer concentration or volatile consumption, public investors may apply a much larger discount than late-stage private investors have.
That does not make the $65 billion figure meaningless.
It makes it the beginning of the valuation debate rather than the end of it.
Anthropic has already answered one of the biggest questions facing generative AI: can frontier models become enormous commercial businesses?
Claude increasingly suggests the answer is yes.
The next question is harder: can those businesses generate durable economics commensurate with trillion-dollar valuations?
That is what Anthropic’s IPO will ultimately test.
