Meta Platforms (NASDAQ:META<\/a>) has spent the year getting punished for the exact strategy that may end up minting money. <\/p>\n Meta runs the largest advertising machine outside of Google, powered by Facebook, Instagram, WhatsApp, Threads, and Messenger, and has quietly become one of the most aggressive infrastructure builders on the planet. Full-year 2025 capex hit $69.7 billion, up from $37.3 billion in 2024, and the 2026 range was pushed to $125 billion to $145 billion. That is Manhattan Project money for GPUs, custom silicon, and data center capacity. <\/p>\n Shares are down 8.7% year to date and 17.4% over the past year, badly lagging the broader market’s advance in 2026. What changed the conversation was a leaked plan showing Meta intends to sell compute externally. That raises the SpaceX comparison bulls have been waiting for. <\/p>\n The SpaceX playbook comes to Menlo Park <\/p>\n SpaceX solved its capex overhang by turning excess capacity into rentable compute. Anthropic agreed to pay $1.25 billion per month for roughly 300 megawatts, and Google\u00a0(NASDAQ:GOOG<\/a>)\u00a0signed a $920 million per month deal for about 110,000 GPUs stretching into mid-2029. That is a $26 billion annual run rate arriving before the S-1 was even dry. <\/p>\n Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Meta didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n Meta has the same ingredients. It owns the buildout, has already signed $107 billion in new contractual commitments this quarter for multiyear cloud deals and infrastructure purchase agreements, and it is deploying more than one gigawatt of custom silicon developed with Broadcom\u00a0(NASDAQ:AVGO<\/a>), alongside a fresh $6.5 billion Samsung foundry deal for its third-generation MTIA accelerator. The core ad engine funds the whole thing. Q1 revenue rose 33% to $56.3 billion at a 41% operating margin, with ad impressions up 19% and price per ad up 12%. Volume and price rising together is rare. <\/p>\n The bear case that keeps working <\/p>\n Story Continues <\/p>\n The bear argument is that this remains a capex black hole. Meta burned 60.2% of its operating cash flow on capex in 2025, Reality Labs is still losing roughly $4 billion per quarter, and the Q1 headline EPS of $10.44 was flattered by an $8.03 billion tax benefit. A single Zuckerberg comment on infrastructure spending sent Applied Optoelectronics down 17% in one session. The market is nervous about ROI slippage. <\/p>\n A depreciation cliff looms. D&A of $18.6 billion trails capex of $69.7 billion by a wide margin. Future earnings absorb a rising drag. Regulatory overhangs in the EU and pending US youth-litigation trials add tail risk that valuation multiples do not always price. <\/p>\n Where patience makes a case <\/p>\n Nobody actually knows if compute-as-a-service materializes into signed contracts this year. Muse Spark is the first model out of Meta Superintelligence Labs, business AI conversations grew from 1 million to 10 million weekly in a single quarter, and yet monetization is still “currently free for most businesses.” <\/p>\n Investors could reasonably wait one or two more prints to see whether third-party revenue arrives before paying up. <\/p>\n What the valuation and ratings show <\/p>\n Meta trades at roughly 21x times trailing earnings and 19x times forward, cheaper than the broader software complex despite 30%-plus revenue growth. The Street consensus target sits at $828.17, or roughly 39.8% upside, on 57 Buy, 6 Hold, and 0 Sell ratings. The ratings distribution is unusually one-sided. <\/p>\n Prediction markets are catching up too. Polymarket assigns 74.5% probability that Meta ends 2026 with a higher valuation than OpenAI, and Deutsche Bank and Morgan Stanley recently flipped their view of Meta’s AI spend from “cash-burning black hole” to “monetization engine.” Meanwhile the stock underperformed the S&P 500 by a wide margin over the past twelve months, which is the setup value investors typically want. <\/p>\n Why $593 is the right entry <\/p>\n At $593, Meta Platforms is a Buy. <\/p>\n The path to price appreciation is bifurcated, and either fork works. If the compute-as-a-service pivot lands even one anchor tenant, Meta reprices as a hyperscaler rather than an ad platform, a multiple expansion story on top of an already-growing earnings base. If it does not, the ad business alone generated $200.97 billion in 2025 revenue at a 41.4% operating margin and continues to compound double digits, which supports the current price without any AI revenue at all. <\/p>\n The entry point matters. Shares sit meaningfully below both the 50-day ($605) and 200-day ($646) moving averages, and the multiple has compressed while earnings have expanded. That is the definition of a re-rating candidate, not an expensive one. The thesis breaks if Reality Labs losses widen materially, if Q2 revenue misses the $58 billion to $61 billion guide, or if promised third-party compute deals fail to materialize by year-end. Those are watchable, not fatal. <\/p>\n The clearest reason to own Meta at this price is that you are getting the ad business at a discount and the AI infrastructure optionality for free. <\/p>\n Act now: the analyst who called NVIDIA in 2010 just named his top 10 AI stocks \u2014 and Meta didn’t make the cut. Grab the names FREE today<\/a>. <\/p>\n<\/a> Chip Somodevilla \/ Getty Images <\/p>\n