Four questions I am forcing myself to answer before I buy it. Steal them for your own tempting stock.

I am bullish on CoreWeave.

I know it is not a quality compounder.

I want to buy it anyway, because I think it can double in a few months.

That sentence is the whole problem. When I catch myself wanting something for a reason that has nothing to do with quality, I stop and write it down. This is that.

I run two systems. The Compounding Portfolio holds 8 to 9 businesses I would own for decades. The Income Portfolio sells premium, never buys it, 45 days out, small and often. CoreWeave fits neither. It is not a decades business. It is not a premium sale. It is a trade wearing an investment costume, and the first thing that should tell anyone is how much conviction they actually have versus how much they want to be in the room while something goes up.

This is what momentum addiction feels like from the inside. Not stupidity. A very good story arriving at exactly the wrong moment for discipline.

So before placing the order, here is the same test recommended to every founder in the Sales Gym before a big deal: find the person who will say no, and listen to them.

The stock, plainly

Picture the chart. CoreWeave went from IPO to $163.66 in twelve months, then gave back nearly half of that in four. It sits at $86.46 now, market cap $47.17B, down 46% from the high, up from a low of $63.80. Revenue is growing at 130% year over year, $6.23B trailing twelve months. It is also losing $1.59B over that same period. Wall Street's consensus is Buy, with an average price target near $132, which prices in real recovery from here.

None of that is fake growth. CoreWeave rents GPUs to the four biggest AI labs on earth: Microsoft, Meta, OpenAI, and now Anthropic. The backlog is close to four times last year's. This is a real business riding a real wave.

It is also carrying real weight.

Four things a skeptic would make you answer

The debt is not decoration. Picture the balance sheet as a second business bolted onto the first: close to $25B in debt at the end of Q1 2026, roughly $7.5B of it secured directly against the GPUs themselves. Analysts expect the covenants on that collateral could trip as early as 2027 if chip values fall faster than the loan assumes. FY2026 capex guidance is $31 to $35B, and the company has already raised $28B in combined equity and debt in the past year. Growth here is not funded by the business. It is funded by the next raise. Say the number out loud, close to $25B, and it does not get quieter the second time.

The depreciation clock and the obsolescence clock disagree. Two clocks are ticking here, and they are not ticking at the same speed. CoreWeave books its GPUs on a six-year useful life, extended from four years back in 2023. The Nvidia architecture cadence that actually determines when a chip stops earning premium rates runs closer to three to four years. Every quarter that gap stays open, reported profitability is doing some of the earnings' work for it. Nebius, a direct competitor, depreciates over four years. That is the more honest number, and it is not the number CoreWeave reports on.

One phone call moves 80% of revenue. Microsoft alone was 67% of 2025 revenue. Even after Anthropic joined the backlog, Microsoft, OpenAI, and Meta together will likely still be more than 80% of 2026 revenue. Meta just announced it is building its own AI cloud infrastructure. The bull answer is that Meta cannot resell CoreWeave's leased capacity through 2032, so the contract holds regardless. That may be true. It also means the entire bull case for the next six years rests on one contract clause holding up against a company with more engineers and more capital than CoreWeave will ever have.

The people with the best information are selling. Founders and executives have sold more than $2.3B in stock since the lockup expired last August. CEO Michael Intrator sold $37.7M on June 30 and $30.4M on June 9. Co-founder Brian Venturo has sold over $1.1B. Insiders selling after a lockup is not automatically a red flag. Insiders selling this much, this consistently, into a stock that is down 46% from its high, is not a coincidence. It is a smell worth staying away from.

A sharper filter: what would Gavin Baker do here

Here is the test worth running on every trade that tempts you: find the investor most fluent in exactly this question, and see where they actually put sized money, not where they put opinions.

Gavin Baker runs Atreides Management and has spent the last two years thinking in public about GPU depreciation, neocloud economics, and semiconductor capex cycles. That is the precise question at hand about CoreWeave.

There is no on-record Gavin Baker quote naming CoreWeave a buy or a pass, and none should be invented. What he has said publicly is more useful than a hot take anyway. At Nvidia's GTC 2026, CoreWeave's own CEO argued that GPU useful life may run 8 to 10 years, not the 4 to 5 the market assumes, because disaggregated inference lets older chips handle prefill and decode work while newer chips handle the frontier tasks. Baker engaged with that argument seriously rather than dismissing it. That is the honest state of the debate: a real, live disagreement between people who understand the hardware, not a settled bear case.

But where Baker has put real, sized money is one layer up the stack from CoreWeave: an aggressive, on-record bet on Nvidia through leveraged call options, increased exposure to memory names SK Hynix, SanDisk, and Micron, and public confidence in TSMC's manufacturing discipline as the thing standing between this cycle and a real glut. Not into the leveraged middleman renting out the chips on borrowed money. That is a pattern, not a quote about CoreWeave, and it is a hard pattern to ignore when the itch is to buy the riskiest layer instead of the ones a specialist actually sized up.

Five scenarios, three to six months out

| Scenario | Price target | Move from $86.46 | What has to happen | |---|---|---|---| | Super Bull | $175 | +102% | A new mega-contract lands (sovereign AI or a fifth hyperscaler), the Meta threat is confirmed contract-locked through 2032, and Aug 11 earnings beat hard enough to reset the depreciation debate in CoreWeave's favor. | | Bull | $135 | +56% | Backlog keeps converting to revenue, customer mix keeps diversifying away from Microsoft, no covenant scare, stock re-rates toward the $132 analyst consensus. | | Base | $90 | +4% | Growth continues near 100%+ YoY, losses persist, Meta overhang lingers as noise rather than a revenue hit, stock chops in a wide range on headline risk. | | Bear | $58 | -33% | A single customer (Microsoft or OpenAI) renegotiates or trims committed capacity, reigniting concentration fear ahead of the 2027 covenant window. | | Super Bear | $34 | -61% | GPU collateral covenants come under real stress, credit markets reprice neocloud debt across the sector the way they hit Nebius and IREN on the Meta news, and CoreWeave is forced into a dilutive emergency raise. |

Notice what that table actually says. The two scenarios that pay well both require something going right that is outside CoreWeave's control: a hyperscaler's capex decision, a contract clause holding under pressure it was never tested against. The two downside scenarios require something that is already visibly in motion: insider selling, a depreciation schedule under scrutiny, a competitor with a trillion-dollar balance sheet entering the market. Base case is a coin flip with a bad payout skew.

What actually happens next

The itch to buy something that can double in a quarter is real. Pretending otherwise would be its own kind of lie.

What changes is the size and the entry. This does not go in the Compounding Portfolio, because it fails the quality test used for every name in it. If it gets taken at all, it is a small, named speculation, sized so that the Super Bear scenario does not touch anything that matters, entered after the August 11 earnings print gives one more real data point on the covenant and customer questions, not before. Conviction that a stock can double is not the same conviction that it is a good business. Being honest about which one you actually have is the whole exercise.

The bitter truth first: the right answer might be a tenth of a position, or none at all. That is the grip discipline is supposed to have on a good story: firm enough to hold, never tight enough to choke the trade.

If you are wrestling with the same itch on your own tempting stock, run it through the same four questions before you place the order: what is the debt actually secured against, does the depreciation schedule match reality, how much of the revenue is one phone call, and are the people with the best information buying or selling. The stock will still be there in the morning. Your capital only gets one shot at compounding correctly.

$50K did not become $4M+ by winning big on stories like this one. It became that by asking these four questions every time a good story showed up wearing an investment costume.

I built the Compounding Portfolio and Income Portfolio systems inside the Sprint Club to stop exactly this: buying a story instead of a business, sized by excitement instead of by rule. Come run your own tempting stock through the same four questions: strategysprints.com

Happy hunting. Simon & The Sprinters 🐬⚡️🐆