Last week, a large retail-focused financial publisher told its readers not to buy Anthropic before its IPO. The stated reason was financial opacity. Within 48 hours, a crypto media outlet had repackaged the same recommendation for its own audience. Two outlets, one argument, zero new data points. That distribution pattern is the signal. When a warning about opacity is itself opaque — republished without a single financial figure attached — the reader should stop asking what the asset is worth and start asking what wrapper they are actually buying. I have audited structures like this before. In 2017, I vetted 14 early token offerings and rejected 11 for undefined tokenomics. The failure mode was never the underlying idea. It was always the wrapper.
Anthropic sits at the front of the foundation-model race. It is structured as a Public Benefit Corporation. It distributes its Claude models through Amazon Bedrock and Google Cloud Vertex AI. It has raised capital across a series of rounds in which Amazon and Google together account for a dominant share of committed funding — Amazon alone has committed roughly $8 billion.
The company is not public. Revenue, gross margin, and burn rate have never been disclosed in full. That is normal for a private company. What is not normal is the speed at which the valuation has been marked up relative to the information available to price it. Public reporting has tracked the mark from roughly $18.4 billion in a 2024 Menlo Ventures round, to about $61.5 billion by early 2025, to market chatter of a far higher figure since. The valuation curve and the disclosure curve are not moving together. That divergence — not the company itself — is the entire substance of the warning.
Retail reaches this story through a narrow channel. Most individuals cannot buy Anthropic shares directly. They buy into a special purpose vehicle that holds a slice of a primary or secondary position, or a platform-wrapped product referencing one. The SPV is the product. The SPV is also the risk.
This structure is not exotic to anyone who has spent time in DeFi. An SPV pools capital, locks it for an undefined period, charges a management fee, and reports a periodic valuation. That is a closed-end fund with a token-like access layer bolted on. We have spent a decade learning how those wrappers fail. Verification precedes valuation; always.
The gap is not that Anthropic is unprofitable — it is that the valuation is being marked up faster than the information required to justify it is being released. A price is a claim about future cash flows. If the claim accelerates and the evidence base does not, the price is carrying an information premium that no buyer can audit. In my own book, I treat any position where I cannot reconstruct the mark from primary data as a position I do not size. That rule is not conservatism. It is arithmetic. If you cannot rebuild the number, you do not own the position. You own a story about the position.
Here is the part the headline skips. The SPV wrapper adds cost layers that never appear in the Anthropic narrative:
- Management and carry. Typical vehicles charge an annual management fee plus carried interest, often in the low-to-mid teens of the gain. The underlying company's performance must clear this hurdle before the investor sees a cent.
- Liquidity lock. Holding periods run for years, frequently with no defined exit. If the IPO window does not open, there is no secondary market and no bid.
- Valuation basis. The reported net asset value is anchored to a specific financing round, not a live market. A stale mark is not a price. It is an opinion with a date printed on it.
- Share authenticity. Not every vehicle can prove clean chain of title to the shares it claims to hold.
In 2022, during the Terra collapse, I ran an emergency withdrawal across three DeFi platforms in 45 minutes and preserved 85% of a €15,000 portfolio — not because I predicted the crash, but because I had pre-verified which contracts I could exit and which I could not. Wrapper risk is verified before the event, never during it.
Contrast this with the 2024 spot Bitcoin ETF. After approval, I ran a statistical arbitrage between spot ETFs and futures, capturing a 120-basis-point spread over three weeks on a €50,000 allocation. That trade worked because the wrapper was public: holdings disclosed daily, fees printed, creation and redemption auditable. The difference between a tradeable wrapper and a trap is not the asset underneath. It is whether the wrapper publishes its own mechanics. Anthropic's retail SPV wrapper does not.
Anthropic's funding is dominated by strategic capital. Amazon and Google are not passive check-writers; they are the company's compute suppliers and distribution channels. When your largest investors are also your infrastructure vendors, the ordinary financial investor sits at the bottom of the information and preference stack. Strategic holders typically negotiate information rights, board observation, and liquidation preference that a retail SPV participant never sees. The retail participant gets the residual — after the strategics are made whole. That is not a conspiracy. It is a capital structure, and capital structures are read, not felt.
Anthropic's Public Benefit Corporation status and its Long-Term Benefit Trust are frequently cited as ethical features. They are also, from an investor-rights standpoint, a structural reason the disclosure picture looks the way it does. A board legally bound to a public-benefit mission does not owe shareholders the same transparency a conventional C-corp board does. That is a deliberate design choice, not a management failure — but it is a design choice external capital must price. Most retail buyers never read it. Before committing to any vehicle of this kind, run the checklist:
- Which financing round anchors the mark, and what is its date?
- What is the full fee load — management, carry, and any platform spread?
- What is the lock-up, and what is the defined exit path?
- Where does this buyer sit in the liquidation waterfall?
- Can the vehicle prove chain of title to the underlying shares?
Five questions. If the seller cannot answer all five in writing, the answer to the investment is no.
Finally, the relay itself matters. A crypto publication repackaging a traditional-finance opinion about an AI company is a cross-domain traffic play, not original reporting. The argument gets compressed. The context gets dropped. What survives is the conclusion — "don't buy" — stripped of the reasoning that produced it. Verification precedes valuation; always. If you cannot trace a claim to its first source, you are not reading analysis. You are reading a summary of a summary.
The obvious reading is that opacity is a scandal. It is not. Opacity is the equilibrium of a competitive race, and treating it as a defect unique to Anthropic mistakes a market structure for a company flaw.
Consider the incentives. In a field where capability gaps between the top labs are narrow and valuations are hypersensitive to perception, publishing detailed unit economics is handing a competitor your pricing model. Every serious lab behaves this way. So a warning framed as "Anthropic is opaque" is misdirected. The transparency problem is real, but it is an industry condition, and it is concentrated in the wrapper layer, not in the company.
There is a second blind spot. The publisher's advice runs opposite to demand. Pre-IPO access to top-tier AI names is scarce, and scarcity is exactly what inflates retail appetite for SPV products. When professional guidance says "do not buy" and retail demand says "let me in," the demand side is the tell — it is running hot. That inversion is the signal worth trading around, not the recommendation itself.
And note the messenger's incentive. A subscription-based advisory that publishes cautionary content has a commercial reason to publish cautionary content. I am not accusing anyone of bad faith. I am applying the same rule I apply to every counterparty: follow the revenue, then weigh the claim.
Do not ask whether Anthropic is a good company. Ask what you are actually buying. Demand the round the mark is based on. Demand the fee schedule, the lock-up, the liquidation preference, and the exit path. If a seller cannot produce those four documents, the answer is no. The next real datapoint is not another opinion — it is a filing. Watch for an S-1, a disclosed revenue figure, or a financing term sheet that names its own valuation basis. Until one of those lands, the honest position on this wrapper is flat — and the discipline that gets you there is the same one that has kept my own book intact through four cycles.
