Before the open, Bloomberg reported that Meta is developing plans for a cloud infrastructure business — internally dubbed "Meta Compute" — to sell access to AI computing power and hosted models to outside customers. CNBC's Jim Cramer confirmed the reporting on air. Shares of Meta (META) jumped ~10% to about $617.69 (from a $563.29 prior close), touching an intraday high of $628.28.
Two paths are reportedly on the table, and Meta may run both in parallel:
The initiative is led by three named executives: infrastructure chief Santosh Janardhan, Meta Superintelligence Labs' Daniel Gross, and president Dina Powell McCormick. Zuckerberg flagged this at the May shareholder meeting as "definitely on the table" if Meta "overbuilt," and said "almost every week" other companies ask to buy access to its compute at a premium. The move puts a name, a team, and an operating plan behind that contingency within roughly two months.
Meta's stock has been a hostage to one number: capital spending. The company guided 2026 capex to $125–145B (raised on higher AI infrastructure costs) and, as of the end of Q1, had committed ~$182.9B to AI infrastructure in coming years — Manhattan-sized data centers in Ohio and Louisiana among them. Industry-wide AI infrastructure spend is projected to exceed $700B in 2026.
That spending crushed the Magnificent 7 in June. The group shed roughly $2 trillion in market cap for the month; the equal-weight Roundhill MAGS ETF fell ~9% (its second-worst month since 2023 launch). Meta specifically fell ~11% in June — the third-worst of the seven, behind Microsoft (−17%) and Amazon (−12%). The Mag 7 are now red for the year, even as the iShares Semiconductor ETF (SOXX) rose ~113% in H1. Apollo's Torsten Slok flagged that free cash flow at Meta, Alphabet, Microsoft and Amazon fell sharply from the 2024 peak. The market has been punishing the check-writers and rewarding the check-receivers.
So a credible plan to recoup cash on idle capacity directly attacks the bear thesis. That is why the stock popped 10% on a report — not a launch, not a revenue number, just a report. The re-rating is about sentiment on the capex hole, not proven cloud economics.
| Metric | Value | Consequence |
|---|---|---|
| Price (intraday Jul 1) | ~$617.69 (+9.7%) | Still −22% below the Aug-2025 high of $796.25 |
| 52-wk range | $520.26 – $796.25 | Bounced hard off the Mar-27 low; today reclaims the pre-June breakdown zone |
| Market cap | ~$1.43T | Cheapest mega-cap ad franchise relative to growth |
| P/E (TTM) | 20.3x | Trailing multiple below the S&P; ad business is not priced for AI upside |
| P/E (forward) | 23.7x | Forward > trailing = the market expects capex/D&A to compress 2026 EPS |
| EPS (TTM) | $27.52 | — |
| Revenue growth (YoY) | +26.2% | Ad engine still compounding >25% — the fundamentals bears forget |
| Gross margin | 81.9% | Software-like margins fund the AI build |
| ROE / ROA | 33.2% / 20.8% | Best-in-class capital efficiency even mid-buildout |
| Beta / Div yield | 1.25 / 0.37% | Growth profile, token dividend |
Fundamentals via Finnhub; prices via Finnhub real-time quote, ~12:00 pm ET Jul 1, 2026.
Meta Compute drops Meta into a market dominated by AWS, Microsoft Azure, Google Cloud, and CoreWeave. But the near-term damage isn't to the incumbents — it's to the pure-play "neoclouds":
| Name | Move today | Why |
|---|---|---|
| CoreWeave (CRWV) | ~$86.82, −12.8% | Its entire model is renting raw GPU capacity — exactly what Meta may now sell |
| Nebius (NBIS) | ~$237.54, −14.0% | Same commoditization threat; high-beta (β ~3) so it moves double |
| Oracle (ORCL) | ~$145.38, −0.8% | OCI competitor but diversified; our existing short still working toward $130 |
| Nvidia (NVDA) | ~$197.20, −1.4% | More rentable supply is marginally negative for GPU scarcity narrative |
| MSFT / AMZN / GOOGL | +3.5% / +1.7% / +0.6% | New competitor, but incumbents shrugged; MSFT up despite 5,500 job cuts |
Meta is CoreWeave's and Nebius's biggest customer. Meta signed a ~$21B expanded agreement with CoreWeave (through Dec 2032) and a deal worth up to ~$27B with Nebius — roughly $48B of committed neocloud spend. If Meta is building its own cloud, the third-order question is: how long before Meta stops renting from the neoclouds and starts competing with them? The selloff isn't just "new competitor" — it's "our anchor customer may become our rival." That's a backlog-quality problem, not a headline-sentiment problem.
Meta is following Elon Musk's playbook. Via xAI, SpaceX began selling excess compute this year — signing Anthropic to $1.25B/month for capacity at its Colossus 1 data center, and Google to $920M/month. The signal, per TechCrunch: the winners of the AI race may not be the best-model builders but the data-center owners — if compute demand holds and data centers retain their value. Both are real "ifs": skeptics warn the buildout leans on rapidly depreciating chips, and the end-user revenue to justify trillion-dollar bets is unproven.
| Scenario | Prob. | Trigger | Price |
|---|---|---|---|
| Bull | 35% | Q2 (late July) quantifies Meta Compute with a signed anchor customer + ad rev holds >22%; 10Y drifts back under 4.30% | $760–800 |
| Base | 45% | Cloud stays "in development," ad engine compounds mid-20s%, capex guide held (not raised again). Stock grinds with the tape | $650–700 |
| Bear | 20% | Q2 capex raised again toward $160B+ with no monetization detail, or ad growth decelerates <15%; hot NFP → 10Y >4.55% | $500–540 |
BUY — HIGH CONVICTION META (long, core)
AVOID / FADE CoreWeave (CRWV) & Nebius (NBIS) dip-buying
WATCH Macro fulcrum — June NFP (Thu Jul 2)
ADP came in light today at 98k (vs 110k est, 122k prior; 48k of it healthcare). NFP consensus is 115k (vs 172k May), UE 4.3%. A cool print keeps the 10Y under 4.50% and is a green light for Meta and duration tech. A hot print that pushes the 10Y >4.55% is the single fastest way to un-do today's pop. Position size accordingly into the binary; Friday is closed.
Meta is the cleanest way to own the AI infrastructure boom while paying almost nothing for the AI upside. Today's 10% pop is a sentiment relief rally on a Bloomberg report, so don't chase it — but the underlying signal (Meta can monetize its build) is real and durable. I'm a buyer into $560–600 with a 6–12 month path to $700–760. The mirror image of that trade is the neoclouds: when your biggest customer becomes your competitor selling the same commodity, "cheap after −13%" is a trap, not a bargain. Own the landlord that owns the demand. Fade the tenant that just lost its anchor lease.
Published by @dailyanalysts on Jul 1, 2026. Opinions are the author's own and are clearly marked. Not investment advice.