Anthropic’s internal 2028 revenue forecast is now the fulcrum for its IPO valuation talks, and it exposes how fragile private AI valuation math can be when it leans on fast-moving run‑rate figures. Bankers are said to be anchoring on a roughly $190–$200 billion revenue outcome for 2028, a number that can justify near‑trillion enterprise values only if today’s headline growth translates into durable, high‑margin sales four years out.
The timing matters because the company’s reported revenue run‑rate has rocketed this year. Anthropic said its run‑rate reached $14 billion by February 12, 2026 and had grown more than 10x over the prior three years (company statement). An investor release tied to its late‑May Series H put the run‑rate at about $47 billion earlier that month and a post‑money valuation near $965 billion (GIC announcement). By mid‑August, media reports citing public‑market checks said the annualized run‑rate had surpassed roughly $65 billion as IPO preparations accelerated (Axios, summarizing Bloomberg).
At the same time, the headline ARR has been disputed. A leaked memo from OpenAI’s chief revenue officer, reported by major outlets, accused Anthropic of “grossing up” cloud‑marketplace sales and overstating its run‑rate by about $8 billion, a disagreement that goes straight to revenue‑multiple math (Reuters reporting via Investing.com). Sector analysts have highlighted how applying public‑market style revenue multiples to private AI labs produces very different answers depending on whether investors accept the headline ARR or insist on a GAAP/net comparable and normalized margins (ValueAddVC analysis).
IPO talks pull Anthropic’s 2028 model into the spotlight
Reuters reported that Anthropic’s internal model projects roughly $190–$200 billion in revenue for 2028 and that this forecast is being discussed in IPO valuation negotiations (Reuters). That use of a longer‑dated revenue number is the material shift. The near‑term run‑rate is climbing quickly, but it is the 2028 figure that, if accepted, can compress forward multiples and support premium pricing on an accelerated listing timeline.
There is, however, dispersion inside Anthropic’s own planning. The Information published alternative internal scenarios: one with about $70 billion of 2028 revenue and around $17 billion of free cash flow, and a lower case with roughly $32.5 billion of revenue and about $3.6 billion of free cash flow (The Information). The range underscores how sensitive the outcome is to growth, pricing, and cost curves.
The strongest data: run‑rate surge vs. 2028 targets
Across primary disclosures and reputable reporting, the picture is of exceptional top‑line momentum paired with contested measurement and divergent forward scenarios.
When
Metric
Figure
Source
Feb 12, 2026
Run‑rate revenue
$14B
Anthropic
May 2026
Run‑rate revenue
~$47B
GIC
Mid‑Aug 2026
Run‑rate revenue
~$65B
Axios
2028 (IPO anchor)
Revenue projection
~$190–$200B
Reuters
2028 (alt scenarios)
Revenue; FCF
~$70B; ~$17B FCF (alt: ~$32.5B; ~$3.6B FCF)
The Information
One additional variable sits underneath these figures: whether run‑rate counts gross cloud‑marketplace billings or net revenue to Anthropic. OpenAI’s leaked memo alleged around an $8 billion overstatement tied to “gross‑up” accounting, though Anthropic has not publicly provided a line‑item reconciliation in response in the cited coverage (Reuters via Investing.com). Analysts have emphasized that valuation outcomes swing widely when investors normalize for such accounting and for eventual margin structure (ValueAddVC).
What the multiples look like if you trust the top line
On the facts, the company’s post‑money valuation was about $965 billion at the Series H close in late May 2026, when investors also pointed to a run‑rate of roughly $47 billion earlier that month (GIC). Simple arithmetic puts that at roughly a 20x annualized run‑rate multiple at that time. If one instead uses the mid‑August reported run‑rate of about $65 billion, the implied multiple compresses materially into the mid‑teens (Axios).
Anchoring on a 2028 revenue figure near $190–$200 billion, as reported in IPO talks, would compress a forward revenue multiple even further, which is why such a forecast is attractive in valuation negotiations (Reuters). But that only works if investors accept that headline revenue will translate into cash generative sales at scale. The Information’s scenarios introduce a much lower outcome set for 2028 revenues and free cash flow, which, if closer to reality, would leave valuation multiples far richer by comparison (The Information).
Accounting mechanics are the swing factor. If cloud‑marketplace pass‑throughs are recorded gross, then a portion of “revenue” may carry reseller‑like economics rather than software‑like margins. As sector analysts note, private‑market comps that do not adjust for this risk comparing apples to oranges when they map AI labs to public‑company multiples that are built on net revenue and consistent gross margins (ValueAddVC). In short, the multiple you think you are paying depends on the accounting you believe.
How this ripples across AI financing and comps
For the company, using a 2028 top‑line anchor could lower the apparent forward multiple and support a larger float at IPO. For investors, the same anchor raises execution risk. In our view, the market will demand a clear reconciliation from gross ARR to a GAAP or net measure, plus a margin roadmap that separates model usage revenue from any pass‑through infrastructure components. Without that, the premium that rapid ARR growth currently commands may fade as diligence tightens.
Across the sector, Anthropic’s framing sets a precedent. Other private AI labs trying to price equity on annualized run‑rates may find that public‑market investors push for net revenue definitions and gross‑to‑net bridges, especially where cloud marketplaces or bundled credits are involved. The episode could also influence how cloud partners structure marketplace arrangements, since revenue presentation affects perceived scale and valuation outcomes. None of this says growth will slow. It does say the multiple could be earned or lost in the footnotes.
Header image from Anthropic’s Feb 12, 2026 Series G announcement (company site) — useful as a primary visual for the firm’s investor messaging and run‑rate claims. — Source: Anthropic (news page image from Series G announcement)
The strongest challenge: a lower 2028 and adjusted ARR
The clearest counterargument is internal dispersion. The Information’s reporting on Anthropic’s alternative 2028 scenarios, including a ~$70 billion revenue and ~$17 billion free cash flow case and an even lower ~$32.5 billion and ~$3.6 billion free cash flow case, suggests the company itself is modeling very different futures (The Information). If either of those paths proves closer to reality than the ~$190–$200 billion revenue forecast, today’s valuation logic weakens.
Second, the ARR dispute matters. If the leaked OpenAI memo’s claim of an ~$8 billion “gross‑up” holds, then any revenue‑based multiple that relies on headline run‑rate would be higher than it appears, and public‑market comps will likely insist on net figures (Reuters via Investing.com). Finally, execution risks remain implicit: sustaining hypergrowth while improving margins requires continued model performance, customer expansion, and cost discipline. Those are reasonable inferences, not verified facts, and they cut both ways.
What would confirm or weaken this valuation thesis
Several concrete disclosures and market signals can validate or challenge the use of a 2028 revenue anchor in Anthropic’s IPO valuation:
- S‑1 revenue policy: explicit revenue recognition language, gross‑to‑net reconciliation for cloud‑marketplace sales, and any principal vs agent determinations.
- Gross margin mix: segment‑level margins that separate core model/API revenue from any pass‑through compute or credits.
- Customer concentration and cohort metrics: top‑customer share, dollar‑based net retention, and usage‑based expansion rates that evidence durability of run‑rate.
- Backlog and RPO: contracted revenue visibility that bridges annualized run‑rate to recognized GAAP revenue.
- Cash flow and capex commitments: free cash flow trajectory and compute purchase obligations that test whether scale begets operating leverage.
- Quarterly cadence: updates that confirm the mid‑August run‑rate level and its progression, relative to the February and May milestones already disclosed.
- Third‑party verification: auditor wording and any independent attestations that reduce uncertainty around the ARR dispute.
Our conclusion: the 2028 forecast is a useful stress test for what premium AI valuations require in accounting clarity and margin proof. Until audited disclosures resolve the run‑rate definition and the margin path, investors should treat that forecast as a scenario, not a foregone anchor.
Disclaimer: This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.
