NEW YORK, July 9, 2026, 13:05 EDT
- Meta flipped higher around midday after opening lower. The stock last traded at about $613.19, up 1.7%. It hit a low of $577.67 earlier.
- Meta aims to put its Iris AI chip into production in September and wants to double its computing capacity to 14 gigawatts in 2027, Reuters reported.
- Investors aren’t just watching AI spend at Meta now. The big question is if every extra gigawatt used can help cut future need for external chip vendors.
Meta Platforms, Inc. NASDAQ:META reversed an early slide Thursday, with shares last at $613.19, up around 1.7%. The stock had dropped as much as 4.2% from Wednesday’s close before turning higher. Investors appeared to set aside worries about the company’s AI-chip spending report.
Traders are now looking at Meta through a different lens, with focus on power, memory supply, custom silicon, and maybe selling extra computing strength. Reuters said an internal memo points to Meta planning to put its own Iris AI chip into production in September, with testing wrapped up in about six weeks and no big issues. Meta wants to boost computing power from 7 gigawatts this year to 14 gigawatts in 2027. A gigawatt measures power and is used in AI data centers to describe how much electricity servers can draw.
Alberta’s build barely drew notice. Meta said this week its first Canadian data center will open with 1 gigawatt of capacity, with plans to scale to 1.8 gigawatts and over C$13 billion committed. That initial phase is 14% of Meta’s planned 7-gigawatt capacity expansion next year; full build-out would be 26%. It’s not a straight cost-per-chip read, but investors can use it as a guide for how capital-heavy these projects are.
| AI buildout marker | Latest disclosed or reported figure | Investor read-through |
|---|---|---|
| 2026 computing infrastructure plan | 7 GW | Meta uses this as its baseline for the AI ramp |
| 2027 target | 14 GW | Doubling goal set for 2027 |
| Alberta first phase | 1 GW | About one-seventh of what Meta aims to add next year |
| Alberta full scale | 1.8 GW | This is close to a quarter of 2027’s planned gain |
| 2026 capex guide | $125 bln-$145 bln | Heavy spending still hangs over the valuation case |
Trading was mixed. Meta outpaced the communication-services ETF, while Alphabet Inc. NASDAQ:GOOGL, its main rival in cloud and ads, traded lower. NVIDIA Corp. NASDAQ:NVDA dropped, keeping its lead in AI chips. Broadcom Inc. NASDAQ:AVGO gained after Reuters reported it’s working with Meta on Iris.
| Security | Recent move | Why investors care |
|---|---|---|
| Meta Platforms, Inc. NASDAQ:META | $613.19, +1.7% | Bounced back as early AI spend worries faded |
| SPDR S&P 500 ETF Trust NYSEARCA:SPY | +0.7% | Main U.S. stock market tracker |
| Invesco QQQ Trust NASDAQ:QQQ | +1.5% | Leans toward Nasdaq tech and growth |
| Communication Services Select Sector SPDR Fund (NYSEARCA:XLC) | +0.2% | Meta’s group trailed the surge in tech heavyweights |
| Alphabet Inc. NASDAQ:GOOGL | -1.9% | Ads and AI peer pulled lower |
| NVIDIA Corp. NASDAQ:NVDA / Broadcom Inc. NASDAQ:AVGO | -1.2% / +3.7% | Nvidia felt GPU risks, Broadcom gained ground with custom chip angle |
The supplier story helped push the shares. Reuters reported Iris is supposed to supplement the GPUs Meta uses for AI and that Meta is working with Broadcom and Taiwan Semiconductor Manufacturing Co. NYSE:TSM, as the company tries to lean less on suppliers like Nvidia. The memo also noted that using the latest GPUs at Meta’s size “has been a heavy lift,” which may explain why investors shrugged off the first move. Reuters
The overall market saw some trouble. Michael Hewson, senior market analyst at iFOREX, said the S&P 500 and Nasdaq put up “a very strong first half,” thanks mostly to gains in memory-storage providers. But Hewson warned profit durability is in question and new tension in the Middle East is creating a “toxic combination.” Since memory chips power the data needs of AI, the comment draws a direct line to Meta’s cost structure. Reuters
Monetization is still the other focus. Reuters, citing Bloomberg News, reported last week that Meta plans to sell extra AI computing power through a new cloud business. Gil Luria, managing director at D.A. Davidson, told Reuters the move would affect “neoclouds” more than major hyperscale providers. Neoclouds only rent out AI compute, not full cloud service suites. Reuters
Meta’s ad business still covers most of the costs. For the first quarter, ad revenue was $55.02 billion on $56.31 billion in total revenue. Ad impressions jumped 19%, with average price per ad up 12%. Capital spending came in at $19.84 billion, free cash flow at $12.39 billion, so AI spending is already high enough to compete with cash generation in investor models.
But the downside is clear. Meta lifted its 2026 capex target to $125 billion-$145 billion, pointing to pricier components and extra data center spending. It also flagged legal and regulatory risks, especially around youth, that could hit results. If Iris misses or Meta’s custom silicon doesn’t ease GPU and memory costs, Thursday’s bounce could end up just a relief rally, not evidence of AI growth.
Next up is the revenue test against rising power costs. Meta has put out a second-quarter revenue outlook of $58 billion to $61 billion. But lately, the stock is trading on whether adding each new gigawatt does enough—more ad efficiency, extra cloud dollars, or cheaper chips—before depreciation ramps up.
Related articles: NASDAQ:META
