The best defense is a good offense. That old sports adage seems to capture what Meta Platforms CEO Mark Zuckerberg hopes to accomplish by standing up a cloud business. Shares of the Facebook and Instagram parent company soared nearly 9% last Wednesday, the day the company confirmed to Jim Cramer that a long-talked-about cloud service was in the works. According to Bloomberg, which was first to report the news, Meta is debating whether to offer access to AI models hosted on its infrastructure or to sell access to raw computing power. A week later, there are few details on exactly what Meta is planning, and the stock is a few dollars per share lower than where it closed the day the news came out. The bull case for a Meta cloud business is that by going on the offensive and creating a new revenue stream, it also provides the defense needed to hedge against overbuilding on AI infrastructure. Those spending concerns were evident on the evening of April 29 when Meta’s outstanding fiscal 2026 first quarter was met with heavy selling. The Street graded Meta’s capital expenditure intentions more harshly relative to its hyperscaler peers ( Amazon , Microsoft , and Alphabet ), as monetization of that spend would depend almost entirely on internal demand for compute. The three biggest public clouds use only the compute they need and rent out any excess capacity. The bear case questions whether a Meta cloud signals excess compute because the company overshot what it needs. Let’s explore each argument and where we come down. The bear case The bear case centers on whether Meta will see enough of a return on invested capital (ROIC) for its massive amounts of AI spending it has already made and the plan for between $125 billion and $145 billion of capital expenditures in fiscal 2026. That’s a capex increase from the prior range of between $115 billion and $135 billion, and above the $122.64 billion expected, even on the low end, according to FactSet. Laura Martin, the longtime Needham analyst, put out a note Monday, saying Meta is entering the cloud business because it overbuilt its AI infrastructure and won’t need all the compute generated by its 2026 capex plans. Martin and her team also said Meta will find the cloud business “difficult to enter this late, owing to well-entrenched deep-pocketed competitors including AWS, Google Cloud and MSFT Azure.” They also see a return on invested capital (ROIC) as Meta “pivots from its 70% margin core advertising business to the 35% margin cloud business.” Meta has been flooding the zone with all kinds of AI announcements, including Tuesday’s release of Muse Image, a new AI model for creating images aimed at attracting creators and advertisers. We noted last month that all these initiatives, from lower-cost smart glasses to an enterprise tool for businesses to plans to build a prediction-markets app and a key partnership with Qualcomm , have done little to get Meta shares out of their funk. META YTD mountain Meta Platforms YTD Meta is the second-worst stock performer year-to-date among its hyperscaler peers, down more than 8%. Microsoft, which is having its own issues around software disruption fears, is down about 20%, while Amazon is up nearly 5%. Alphabet has been the big winner thus far, gaining over 15% year-to-date. Meta is also trading at the lowest multiple — at 17.7 times next 12 months’ earnings estimates, versus Amazon at 25.5 times, Microsoft at 19.6 times, and Alphabet at 24.9 times. Canaccord Genuity analysts think the Meta bear case has “gone too far.” They said in a note on Monday: “With an accelerating ad business, emerging subscription tiers, and an external market now paying observable rates for capacity, META’s discount to the rest of the Mag 7 looks increasingly difficult to justify.” The bull case That brings us to the bull case. JPMorgan estimated that every gigawatt of Meta compute capacity offered in a cloud business could generate $20 billion of annual revenue and several dollars of earnings per share (EPS). Meta’s revenue in fiscal 2025 rose 22% to nearly $201 billion. EPS for fiscal 2025 slipped 1.6% to $23.49. (Compute is measured in gigawatts because power is the limiting factor in data centers. So, 1GW of compute means AI infrastructure that draws 1GW of continuous energy, or enough to power up to 1 million homes.) Meta has made it known that — for now — it needs all the compute it can get. While a cloud business may result in an even greater internal capacity constraint, the move provides the hedge the company needs to rationalize its spending. We already know that these tech CEOs think the risk of underspending outweighs the risk of overspending — and as a result, they are willing to take on the risk of overspending. The cloud plan for Meta provides an off-ramp should the day come that Meta is sitting on a few too many gigawatts of excess capacity. In addition, Meta can significantly expand demand for that compute as it opens its infrastructure to the world. That demand will also be more diverse because it will come from many different companies across many industries, each in a different phase of the business cycle and with its own unique AI demand profile. This is why the public cloud business model is so successful. Club view As Meta investors for the Club portfolio, we need to consider both sides of the debate. But perhaps it does not have to be one or the other. Maybe Meta is indeed still compute-constrained internally. It can also be true that Zuckerberg determined that if he wants Wall Street behind him and the stock, then he has to provide a rock-solid reason for all the spending. A public cloud business achieves both goals simultaneously. Why is that an attractive proposition? Consider what we just learned about Meta’s relationship with Alphabet. The Google parent told Meta it would need to limit the use of Gemini because it didn’t have the capacity to meet Meta’s demand. By being the renter rather than the landlord, Meta is at the mercy of those with compute. When Alphabet decides it needs more internally, it can limit supply to its customers. Wall Street doesn’t appear to look favorably on companies that have the option to rent out compute, but instead hoard it for internal initiatives — especially when there isn’t much confidence that the internal plans can yield as great a return, or as near-term a return as simply renting it out. That’s part of the reason for Microsoft’s dismal performance this year. (Microsoft also has too much exposure to enterprise software, which, as a group, has been bludgeoned by AI disruption worries.) For Meta, the company is more likely to reallocate any internal compute that is ultimately deemed to yield a lower return and rent it out — in essence, doing the opposite of what some investors have taken issue with at Microsoft. That should please the Street. We view it as a strong move because it plays offense while doing wonders in defending against the dizzying levels of capex needed to ensure it is not left behind. (Jim Cramer’s Charitable Trust is long META, AMZN, MSFT, GOOGL. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
