💻 Computing
Anthropic Is Projecting $200 Billion in 2028 Revenue. Google Took 22 Years to Get There.
Anthropic's bankers are telling prospective IPO investors that the company will generate $190 to $200 billion in revenue by 2028, a figure the company itself projected at $70 billion just nine months ago. We built the revenue velocity comparison that nobody has run and found that Anthropic, if it hits these numbers, will have traveled from first commercial dollar to $200 billion roughly 4.5 times faster than Google, the previous record holder among technology companies.
Nine months ago, Anthropic told investors it expected $70 billion in 2028 revenue. That was November 2025. The company's annualized run rate at the time was $9 billion, already absurd for a four-year-old startup. Seventy billion felt ambitious. Then the number kept climbing. By April 2026 the run rate hit $30 billion. By May it crossed $47 billion. And on Friday, Reuters reported that Anthropic's bankers are now pitching prospective IPO investors on 2028 revenue of $190 to $200 billion, roughly triple the figure the company itself provided less than a year earlier.
That is not a rounding error. It is a 171 to 186 percent upward revision in under twelve months. Revenue projections that nearly triple while the company is simultaneously preparing for a $2 trillion IPO are, to put it gently, unusual.
The Velocity Comparison Nobody Ran
The raw dollar figure is staggering, but what makes it historically anomalous is the speed. We assembled revenue trajectory data for every company that has reached or approached $200 billion in annual revenue, measuring from each company's first year of material commercial revenue to the year it crossed the $200 billion threshold.
| Company | First ~$1B Revenue | Year at ~$200B | Time |
|---|---|---|---|
| Walmart | ~1979 | ~2001 | ~22 years |
| Apple | ~1983 ($1B) | ~2020 ($274.5B) | ~37 years* |
| Amazon | ~1999 | ~2018 ($232.9B) | ~19 years |
| Alphabet/Google | ~2003 ($1.47B) | ~2021 ($257.6B) | ~18 years |
| Meta/Facebook | ~2009 | Not yet reached | N/A |
| Anthropic (projected) | ~2024 ($1B ARR) | 2028 ($190-200B) | ~4 years |
Google generated its first billion dollars in 2003. It did not cross $200 billion in annual revenue until 2021. Eighteen years. Amazon hit its first billion in 1999 and needed until 2018 to reach $200 billion. Nineteen years. Apple technically crossed $1 billion in 1983, nearly went bankrupt in 1997, and did not reach $200 billion until 2020, a 37-year span interrupted by a near-death experience that makes the comparison imperfect. Anthropic is projecting to cover the same distance from $1 billion in annualized revenue, which it reached at the start of 2025, to $200 billion by 2028. Four years. Even measured against Google, the cleanest comparison, that compression ratio of 4.5x has no precedent in the history of corporate revenue growth.
The Projection Itself Is the Story
What makes this different from a typical startup forecast is the trajectory of the forecast itself. Most companies revise revenue projections upward by 10 to 20 percent as execution improves. Anthropic revised its 2028 target by 171 percent in nine months. Here is the sequence:
- October 2025: Reuters reports Anthropic projects $20 to $26 billion in 2026 annualized revenue
- November 2025: The Information reports $70 billion in 2028 revenue, $17 billion in 2028 cash flow
- April 2026: Run rate crosses $30 billion, CEO Dario Amodei says he has "consistently underestimated" his own growth
- May 2026: Run rate reaches $47 billion, Series H closes at $965 billion valuation
- August 2026: Bankers pitch $190 to $200 billion for 2028
From $87 million in annualized revenue in January 2024 to $47 billion by May 2026. That is a 540x increase in 28 months. For context, Google grew approximately 3x annually during its fastest growth phase from 2002 to 2007. Anthropic grew 540x in a period shorter than a single presidential term.
Where the Money Comes From
The revenue composition matters because it determines whether $200 billion is plausible or fantastical. Anthropic is not a consumer app selling $20 monthly subscriptions. Roughly 80 percent of its revenue comes from enterprise API customers paying per token for Claude model access. Over 1,000 businesses now spend more than $1 million annually, a figure that doubled in less than two months during early 2026. Claude Code, the company's agentic coding product, alone reached $2.5 billion in annualized revenue by February 2026.
This is the structural difference between Anthropic and OpenAI. OpenAI generates a large share of its revenue from ChatGPT consumer subscriptions. Anthropic generates the vast majority from enterprise API spend billed by the token, which scales with corporate deployment depth rather than individual user conversion. When Goldman Sachs or Amazon Web Services doubles the number of Claude API calls their internal tools make, Anthropic's revenue doubles from those accounts without a single new customer signing up.
Capital Efficiency: The Divergence With OpenAI
Both companies sell access to frontier AI models. Both spend billions on compute. Both are preparing for IPOs that could value them above a trillion dollars. But their financial architectures are diverging in a way that the $200 billion projection makes impossible to ignore.
| Metric | Anthropic | OpenAI |
|---|---|---|
| Current run-rate revenue | $47B (May 2026) | ~$24-25B (Feb 2026) |
| 2028 revenue projection | $190-200B | $100B (2027 target) |
| Total funding raised | ~$87B | ~$150B+ |
| Projected cash flow positive | 2028 ($17B) | Not before 2029 |
| Cumulative cash burn | Not disclosed | $115B through 2029 |
| Compute cost per revenue $ | $0.56 (Q2 2026E) | Higher (consumer-heavy) |
| Enterprise customers ($1M+) | 1,000+ | Not disclosed |
| Revenue mix | ~80% enterprise API | Mixed consumer/enterprise |
Anthropic has raised roughly $87 billion across nine funding rounds. If it generates $200 billion in 2028 revenue, its revenue-to-cumulative-funding ratio would be 2.3x. That number sounds reasonable until you look at OpenAI: the company has raised over $150 billion, targets $100 billion in 2027 revenue, and projects cumulative cash burn of $115 billion through 2029, meaning it will have spent more money losing money than Anthropic will have raised in total. Anthropic projects positive cash flow at $200 billion revenue. OpenAI projects continued losses at $100 billion revenue. Both companies sell functionally similar products to overlapping customer bases. Something in these projections cannot be simultaneously true.
The Implied Market Size Problem
If Anthropic generates $200 billion from AI model access and holds approximately 35 percent of the enterprise API market, based on Ramp's May 2026 AI spending data showing Anthropic at 34.4 percent of enterprise AI corporate-card spend, the implied total AI model API market is approximately $570 billion by 2028. Add cloud AI services from AWS, Azure, and Google Cloud at another $150 to $200 billion. The total AI revenue market implied by Anthropic's projection is $720 to $770 billion.
Global GDP is roughly $115 trillion. A $750 billion AI market would represent 0.65 percent of global economic output. For perspective: the entire global airline industry generates about 0.4 percent of GDP. The entire global advertising market is approximately $1 trillion. Anthropic's own revenue projection, if true, implies that AI revenue will reach 70 percent the size of global advertising within two years. That is either the most consequential economic transition since the internet or the most consequential forecasting error since WeWork.
The 2028 Forward Multiple
Anthropic's expected IPO valuation of $2 trillion divided by its projected 2028 revenue of $200 billion produces a forward multiple of 10x. In isolation this sounds reasonable: high-growth SaaS companies routinely IPO at 10 to 20x next-twelve-months revenue. But this is not next-twelve-months. It is a 10x multiple applied to revenue projected two full years into the future. Reuters notes this approach has precedent: Cerebras and SpaceX both used extended forward projections in their recent IPOs. SpaceX went public at $1.8 trillion in June 2026 using projections that extended to 2029.
If Anthropic hits the $200 billion number, the $2 trillion IPO will look cheap at 10x. If the number revises downward, as projections that triple in nine months sometimes do, the multiple re-rates brutally. A company valued at $2 trillion on the basis of $200 billion future revenue that delivers $100 billion instead is suddenly trading at 20x, expensive by any standard, and the investor who bought the IPO at $2 trillion finds themselves holding an overpriced position in a company that merely doubled its revenue from $47 billion to $100 billion, a result that would be remarkable for any other company in history and disappointing only because the promise was twice as large.
The Strongest Counterargument
The strongest case against the $200 billion projection is that it requires Anthropic to grow from $47 billion in annualized run-rate revenue to $200 billion in roughly 2.5 years, a 4.3x increase. This assumes enterprise AI budgets scale proportionally, but CIOs eventually hit budget ceilings. Corporate IT spending globally is about $5 trillion annually. If Anthropic captures $200 billion, that is 4 percent of all corporate IT spending worldwide flowing to a single AI vendor, a concentration that has never existed in the history of enterprise technology, not for Microsoft, not for Oracle, not for Salesforce.
Pricing pressure is the more immediate threat. Anthropic charges $3 to $15 per million tokens depending on the model. Google offers comparable models with Gemini, bundled into existing enterprise relationships through Google Cloud. Meta releases open-source models that enterprises can run themselves. If a price war erupts, every dollar of Anthropic's per-token revenue is at risk, and a 50 percent price decline would make $200 billion unreachable at any volume that current infrastructure could serve.
Volatility of the projection itself is a warning sign. A 2028 revenue estimate that moved from $70 billion to $200 billion in nine months, a 186 percent increase, could just as easily move in the other direction. Projections that triple during the pre-IPO roadshow period are projections optimized for fundraising, not necessarily for accuracy.
Limitations
Anthropic is a private company. Every revenue figure cited in this analysis comes from secondhand reporting: Reuters, The Information, Bloomberg, Sacra, and public statements by investors like Brad Gerstner. We have no access to Anthropic's actual financial statements. Sacra's $47 billion May 2026 run-rate estimate is not an Anthropic disclosure. Run-rate revenue, which annualizes the most recent monthly or quarterly figure, overstates actual annual revenue if growth decelerates or understates it if growth accelerates. Our market share estimate of 35 percent relies on Ramp's corporate-card spending data, which may not be representative of all enterprise AI spend. The historical velocity comparison uses approximate dates for "first material revenue," which is inherently subjective. OpenAI's projected $115 billion cumulative cash burn comes from The Information reporting on internal financial plans, not from public filings.
What You Can Do
If you allocate enterprise AI budgets: lock in pricing now. Anthropic's current token pricing reflects a company that needs to grow from $47 billion to $200 billion. Volume discounts negotiated today will be harder to extract once the company is public and managing quarterly earnings expectations. Ask your account team for multi-year rate commitments, in writing, before the IPO.
If you evaluate AI investments: do not compare Anthropic's $2 trillion IPO valuation to its current $47 billion run rate. Compare it to the $200 billion 2028 projection, which is how the bankers are pricing the deal, and then ask yourself what probability you assign to the projection after watching it triple in nine months. A 10x multiple on a number that moved 186 percent in under a year is a bet on the projection's stability, not just its magnitude.
If you build AI applications on Claude APIs: model two scenarios. In one, Anthropic hits its numbers and raises prices with the pricing power of a near-monopoly enterprise AI vendor. Your compute costs rise 30 to 50 percent. In the other, a price war with Google and open-source alternatives drives per-token costs down 60 percent. Your unit economics improve dramatically but Anthropic's revenue target becomes unreachable, potentially triggering cost-cutting that affects service reliability. Build your architecture to survive both worlds by keeping model-switching costs low, maintaining evaluation benchmarks across providers, and avoiding dependencies on features unique to a single vendor.
The Bottom Line
Anthropic is attempting something that has never happened in the history of commerce: reaching $200 billion in annual revenue within five years of its first commercial dollar. Not five years from founding. Five years from first dollar. Its projection is not a forecast in the traditional sense. It is a bet, made by bankers pricing a $2 trillion IPO, on the assumption that enterprise AI spending will grow faster than any corporate technology category has ever grown, and that a single company will capture enough of that growth to generate more revenue than all but a handful of companies on Earth. The bet might be right. The fact that the same projection was $70 billion nine months ago should give everyone, especially the people writing checks at the IPO, a reason to measure twice.