AI spending keeps growing.
Companies keep increasing what they spend on building and running AI systems, and that money is collected at three different points on the way through.
Saved in this browser, not to an account.
The time-bound call
Fixed at publicationThis basket will outperform SPYx over 90 days.
- Started
- Deadline
- Last observation
How the call is scored
A fixed model basket versus SPYx, each initially quoted from $150 USDC. The original raw token quantities never rebalance. Returns compare their USDC sell quotes with the starting sell quotes. The call succeeds only if the basket's return is strictly higher. Equal returns are a tie. Resolution uses the first complete quote observation started on or after the deadline, within 48 hours. If that window has no valid observation, the result is unresolved. Quotes include route price impact but exclude wallet execution and network costs. These are model returns, not your investment results.
Each quote must have positive output, at most 1% price impact, and a context slot within 450 slots of the current chain. A complete observation must finish within 30 seconds.
Reference observations refresh daily. A deadline scores this model; it never sells your holdings or pays a prediction prize. A new thesis version gets a new call.
Quote source: Jupiter ↗The argument
Why this idea. Why these businesses.
One company sells the hardware the spending buys. One rents it out and sells software on top. One sells the software that turns a model into a decision inside an organisation. The money passes all three, but it arrives at different times and with very different margins.
Inside the basket
- 40%
Microsoft
Microsoft rents out the compute and sells the software that runs on it. It earns whether a customer builds their own system or buys one already built.
The tradeoff. It pays for the capacity up front. The spending lands on the accounts before the revenue does, so a year of heavy building can look worse than a year of standing still.
About this tokenized stock
Underlying: MSFT. Issuer: Backed Finance. Backed Finance issues these tokens. The issuer can move tokens out of any wallet (permanent delegate) and can freeze all transfers (pausable). Balances rebase for dividends and splits, so your share count can change without a trade.
Issuer terms (opens in a new tab) - 35%
NVIDIA
Almost everything being built runs on hardware NVIDIA designs. When spending rises, this is the first place the money lands.
The tradeoff. It is the most crowded way to hold this idea, and its largest customers are designing their own chips to avoid paying it. A supplier with substitutes loses margin before it loses revenue.
About this tokenized stock
Underlying: NVDA. Issuer: Backed Finance. Backed Finance issues these tokens. The issuer can move tokens out of any wallet (permanent delegate) and can freeze all transfers (pausable). Balances rebase for dividends and splits, so your share count can change without a trade.
Issuer terms (opens in a new tab) - 25%
Palantir Technologies
Palantir sells the layer that turns a model into a decision inside a large organisation. It is the part of the spending that has to justify itself to a budget holder.
The tradeoff. Revenue is concentrated in a relatively small number of large contracts, many with governments, so a single procurement cycle moves the business. It is the smallest and least diversified holding here, which is why it carries the smallest weight.
About this tokenized stock
Underlying: PLTR. Issuer: Backed Finance. Backed Finance issues these tokens. The issuer can move tokens out of any wallet (permanent delegate) and can freeze all transfers (pausable). Balances rebase for dividends and splits, so your share count can change without a trade.
Issuer terms (opens in a new tab)
Why these weights. The weights fall as the holding becomes more concentrated on a single expression of the claim. Microsoft's business survives a slower year; Palantir's is the most exposed to one. This is a judgement about concentration, not a forecast about returns.
The strongest case against
This is the thesis most likely to be right about the world and wrong about the investment. Spending is a cost to the spender: capacity bought now shows up as depreciation for years, whether or not the revenue arrives. The supplier's biggest customers are also designing their own chips, and a supplier whose customers have alternatives loses margin first. A world that spends enormously on AI and earns thin returns on it would confirm the claim and still damage every holding here.
What would change the thesis?
A quarter in which the largest cloud buyers guide their capital spending down, or two consecutive quarters where NVIDIA's gross margin falls while revenue still grows. Either would say the spending is being met by supply rather than constrained by it.
Read the evidence. Make up your mind.
Sources supporting the idea, including the ones that challenge it.
- NVIDIA Announces Financial Results for Second Quarter Fiscal 2027 (opens in a new tab)
NVIDIA, 8-K exhibit via SEC EDGAR
The spending arriving at the supplier, and the forward guide rather than only the past: data centre revenue of $89.0 billion, up 117% year-over-year, with next-quarter revenue guided to $108.0 billion.
- Microsoft Cloud and AI Strength Fuels Fourth Quarter Results (opens in a new tab)
Microsoft, 8-K exhibit via SEC EDGAR
The buyer's side of that number, in an audited cash flow statement rather than a press line: $115.9 billion of additions to property and equipment across the fiscal year, with Azure revenue up 43%.
- Amazon.com Announces Second Quarter 2026 Results (opens in a new tab)
Amazon.com, 8-K exhibit via SEC EDGAR
A second buyer spending at the same magnitude, so the first is not an outlier: $169,007 million of property and equipment purchased over twelve months, with AWS at a $169 billion annualised run rate.
- Global Financial Stability Report, April 2026, Chapter 1 (opens in a new tab)
International Monetary Fund
An official body treating this spending as concentration risk rather than as a given. It notes hyperscalers are expected to account for 70 percent of a projected $3.4 trillion of AI capital expenditure by 2029, and warns those valuations may not be justified by returns.
Updates and version history
An update appends dated evidence to this argument. A new version changes the argument or the allocation itself. Neither one touches a position you already hold.
No updates since publication. When the author adds one, it appears here with its date and its sources — it never changes what is written above.
- Version 1
2218e34c4fCurrent
Written by the Thesis team. The team holds no position in these companies or their tokenized shares, and is paid nothing by any of them.
Tracking begins at publication. This thesis has no established performance history. Tokenized stocks carry issuer and market risk.