Wall Street has spent six months asking the wrong question about Meta Platforms Inc (NASDAQ: META). The debate has centered entirely on whether $145 billion in capital expenditures can be justified by an advertising business, and while there’s nothing inherently wrong with that, that framing misses what has been happening in plain sight: a systematic, deliberate construction of the infrastructure, the silicon, the partnerships, and the executive talent required to sell compute to the world. Not as a side project. As a business.
And Meta earnings call on July 29 may be the moment that debate becomes irrelevant.
The Capex Isn’t A Cost But A Down Payment
When Meta published “What Is Compute Power?” on June 10 for a general audience, explaining CPUs, GPUs, FLOPS, and gigawatts in language a first-year student could follow, most people read it as a PR exercise.
I see it as a product announcement without a price tag. Companies don’t publish infrastructure primers for consumers unless they’re building toward something that requires public understanding before they can sell it to them. AWS didn’t explain cloud computing to the general public until it needed the public to buy cloud computing.
The June publication also revealed Meta is building custom MTIA chips in partnership with Broadcom, co-developing Arm’s first data-center processor built specifically for AI workloads, and running partnerships simultaneously with AWS, AMD, and Nvidia to supply the broader compute portfolio.
Four new chip generations deploying within two years. This is beyond the scope of a company buying infrastructure to run Facebook better. Meta is now building a vertically integrated compute stack – custom silicon through data centers through models – that mirrors what Amazon Inc (NASDAQ: AMZN) assembled before AWS became a $100 billion annual revenue business.
Meta is also in early talks to lease $10 billion in compute to Anthropic over two years, its own direct AI competitor. The internal name for the cloud effort is already circulating: Meta Compute. And the company hired Dave Brown, a former senior Amazon Web Services executive with 19 years building AWS’s compute and machine learning services, with a mandate to build that business and a reporting line directly to Meta’s head of infrastructure. You don’t hire an architect of AWS to run a data center cost center.
The advertising business has never been stronger, and the market still sold the stock down nearly 10% year-to-date to $595. Q1 revenue grew 33%, the fastest pace since 2021, ad impressions increased 19%, and average price per ad climbed 12%.
Consensus for Q2 sits at $60.26 billion in revenue and $7.23 in EPS, with Polymarket traders assigning an 87.1% probability of another beat.
To be honest, the ad business hasn’t given investors any reasons to be spooked yet. Rather, investors’ skepticism is about what the $145 billion buys beyond it, and whether EPS growth of just 1.3% year-over-year against 26.8% revenue growth represents structural margin compression or temporary investment cost that eventually generates a second revenue stream.
That’s the uncertainty being priced at 18x forward earnings, cheap for a business growing revenue at 33%, and it’s the uncertainty that resolves if Zuckerberg walks onto the call on July 29 and confirms Meta Compute as a named, funded, publicly committed cloud business rather than a Zuckerberg offhand comment from Q1.
Every Rally Since February Has Failed At The Same Trendline
Meta trades at $605, essentially pinned against its 50-day moving average at $605.84 on volume of 133,660 shares: a quiet, controlled session that tells you institutions are holding rather than distributing ahead of a binary catalyst two days away. The 20-day sits at $623.71 above, the 200-day at $637.53 further above, and the descending trendline drawn from the February high near $740 continues pressing down toward current price. Every meaningful rally attempt since February has failed beneath that trendline, which means the stock has been repricing lower while the business has been accelerating higher… a divergence that either resolves through a fundamental re-rating or confirms that the capex concern wins the argument.
A confirmed Meta Compute announcement on July 29 doesn’t just beat an earnings estimate. It changes the valuation framework entirely. A cloud business valued on infrastructure multiples, where AWS alone contributes over $100 billion annually to Amazon’s revenue, sitting on top of an ad machine already growing at 33% produces a different stock price than an ad company spending aggressively on unproven AI ambitions. The trendline resistance near $680 and the 200-day at $637 become the first levels worth watching if the call delivers confirmation.
If Zuckerberg Names It, The Multiple Has To Change
Revenue and EPS are largely priced. The beat is expected. What I want from Zuckerberg is specificity, not a passing reference to “monetizing excess capacity” but a named business, a revenue timeline, and a clear signal that the Anthropic deal and the Dave Brown hire represent the beginning of a cloud revenue strategy rather than opportunistic infrastructure management.
If that confirmation arrives, Meta stops being an ad company with an expensive AI habit and starts being priced as the only company in the world simultaneously running the largest social advertising network on earth and building the compute infrastructure to lease to everyone else building the future on top of AI.
The Canada data center, the Anthropic talks, the infrastructure explainer, the AWS executive hire, none of these are coincidental. They’re sequential. And July 29 is where the sequence either gets named or gets delayed another quarter.
I’m bullish. The stock at 18x forward earnings for a business with this growth profile and this optionality is mispriced even before Meta Compute enters the conversation. The confirmation just accelerates the timeline for the market to figure that out.
