Arm FQ4 & CoreWeave Q1 – Rocky Road

Arm FQ4 26 – RSI curse.

• Arm is well on its way to claiming a piece of the booming data centre market, but the cautious smartphone commentary on the conference call and the fact that the shares have doubled in the last 6 weeks provided an excuse to take some profit.
• FQ4 26 revenues / Adj-EPS were $1.49bn / $0.60, slightly ahead of consensus, but could have been more if the smartphone market were not suffering.
• This is because the smartphone market is being hurt by the memory shortage, which is mainly affecting the mid-range of the market.
• Here, royalty revenues were $671m in FQ 4 26, which was slightly behind expectations, and the company was fairly sanguine about the short-term outlook for smartphones.
• This was offset by licensing revenue of $819m, well ahead of estimates of $776m, which I take as positive, as this is a good indicator of royalty revenues that will follow as products get to market.
• It is in the longer-term where the good news was to be found as Arm has secured $2bn in orders for this fiscal year and next and is confident of its $15bn in data centre CPU revenues by 2031.
• This is crucially important as this is the scale of revenue that the company needs to hold onto its current valuation, and now there is a pretty good visibility as to how it will get there.
• This is why the shares have doubled over the last 6 weeks, which has led to the shares looking very overbought on the relative strength indicator (RSI).
• I suspect that this is the biggest factor behind the 10% correction after the results, as the weakness in the smartphone market looks to be already in the estimates, as guidance for the coming quarter was broadly in line with expectations.
• Here, FQ1 27 revenues / Adj-EPS will be $1.21bn – $1.31bn ($1.26bn) / $0.36 – $0.44 ($0.40), slightly ahead of consensus at $1.25bn / $0.36.
• This strongly implies that consensus is already including weakness from the smartphone market, further supporting my contention that the shares were overbought in the short-term rather than a sea change in how the market is looking at the company.
• Hence, the story is now looking much better as the company has provided a clear and credible explanation of how it intends to grow revenue, and so I don’t think that this is the beginning of a major correction in the share price.

CoreWeave Q1 26 – Revenue per GW bomb

• CoreWeave reported disappointing results as its forecast did not meet expectations, but the real problem remains that it is not earning nearly enough revenue per GW of capacity that it already has online.
• This is a major red flag, and unless this improves, CoreWeave is going to run out of money, resulting in a likely fire sale and a wipeout of the equity.
• Q1 26 revenues / were $2.1bn / LOSS($1.12) below estimates of $2.0bn / LOSS($0.97) and the outlook did not fare much better.
• Here, Q2 26 revenues are expected to be $2.45bn – $2.60bn ($2.52bn) below expectations of $2.67, which served to further dent optimism.
• To be fair to CoreWeave, its revenue miss is most likely due to delays in building out and turning on its capacity, but it is in the amount of revenue it is generating where I have real concerns.
• The average capacity it had online during Q1 26 was 925MW, upon which it generated $2.1bn, which is equivalent to annual revenues $9.1bn / GW
• RFM Research has established that with the GB300 infrastructure that CoreWeave is using, $10bn / GW is the absolute bare minimum that the company can earn, and even this amount generates a paltry 5-year return of 1% on a cash flow basis.
• This means that CoreWeave urgently needs to increase revenue per GW or face insolvency.
• On this basis, its FY 26 guidance is not encouraging as it expects to generate FY 2026 revenues of $12.5bn with an average deployed capacity of 1.28GW during the year ($9.8bn / GW).
• This means that it expects revenue per GW to remain flat for the year, which, in my opinion, has serious implications for its outlook as a going concern.
• The balance sheet also makes for concerning reading with $17.8bn in long-term debt, $7.5bn in short-term debt, $2.1bn in deferred revenue in the next 12 months, and $5.4bn deferred revenue for periods beyond 12 months.
• Deferred revenue is cash that the company has already received as pre-payments for its products and services.
• This means that cash flow is going to be negatively impacted as these sales are recognised, but no cash is received.
• CoreWeave also expects to spend $31bn – $35bn in capex that will deliver around 800MW of capacity giving a cost per GW of $40bn.
• This is in line with industry estimates, but it is way above what Nebius says it can achieve, where it has guided that it will spend around $25bn / GW of capacity (see here) in 2026.
• This will have a material impact on Nebius financial performance as at $10bn per GW on capex of $25bn / GW, the 5-year cash flow return improves to 16%, which offers Nebius’ backers a good return on their investment.
• Hence, CoreWeave needs to either improve its revenue per GW by at least 20% – 30% or reduce its capex per GW by a similar amount to continue as a going concern.
• This looks unlikely to occur this year, and so I suspect that losses will continue, resulting in indebtedness and equity dilution continuing to grow.
• It is the suppliers of CoreWeave where one wants to have one’s money.