What Happened?
Shares of social network operator Meta Platforms NASDAQ:METAjumped 5.2% in the afternoon session after a cluster of AI infrastructure and product news reframed the company's heavy capital spending as a path to lower costs, new revenue, and faster growth rather than a pure expense drag.
On the cost side, Meta said it will begin manufacturing its custom "Iris" AI chip in September with Broadcom and TSMC, aiming to cut reliance on pricey Nvidia and AMD GPUs. As Meta owns more of the silicon stack, inference and training could get cheaper over time.
Monetization is the second leg. Meta is building a $9.1 billion (CAD 13 billion) data center in Canada and targeting 14GW of total compute by 2027. By renting excess capacity as an "AI Cloud," it turns idle or surplus infrastructure into a revenue stream. Scale also helps on unit economics: Meta's build costs sit below the roughly $45 billion-per-GW industry average, so the same dollars buy more usable capacity than many peers can claim.
Growth closes the loop. Muse Spark 1.1, Meta's advanced coding model, is priced at about one-quarter of comparable OpenAI and Anthropic offerings, a deliberate bid to pull developers into Meta's ecosystem and widen the funnel for tools, ads, and cloud demand later.
The risk is execution. Custom silicon, multi-gigawatt buildouts, and a cut-price model only pay off if utilization, margins, and developer adoption follow the spend. Until those show up in results, markets can still punish CapEx as overbuild rather than investment.
What Is The Market Telling Us
Meta’s shares are somewhat volatile and have had 11 moves greater than 5% over the last year. In that context, today’s move indicates the market considers this news meaningful but not something that would fundamentally change its perception of the business.
The previous big move we wrote about was 9 days ago when the stock gained 10.1% as reports revealed that the company plans to build a cloud business to sell its excess AI computing capacity.
Meta is developing a cloud arm (internally "Meta Compute") to rent out spare capacity, both as Model-as-a-Service (hosting its own Muse Spark models, akin to AWS Bedrock) and raw compute (akin to neoclouds like CoreWeave).
CEO Mark Zuckerberg confirmed at the annual shareholder meeting that a cloud business was "definitely on the table."The central bear case on Meta all year had been the capex bill: 2026 AI capital spending was guided up to $125–145 billion (from $115–135 billion, versus ~$72 billion in 2025), pressuring free cash flow and sinking the stock ~7% after Q1 despite an earnings beat.
A cloud business directly monetizes that infrastructure turning a feared cost center into a potential revenue line and positioning Meta as the fourth US hyperscaler alongside AWS, Azure and Google Cloud.
Meta is up 2.6% since the beginning of the year, but at $667.43 per share, it is still trading 15.5% below its 52-week high of $790 from August 2025. Investors who bought $1,000 worth of Meta’s shares 5 years ago would now be looking at an investment worth $1,890.
ALSO WORTH WATCHING: Nvidia’s Quiet Partner. Nvidia’s chips cost a hundred grand. The connectors that make them work cost even more. One company makes them all.
Every AI server needs specialized infrastructure the chip companies don’t make. High-speed cables. Power connectors. Thermal sensors. This 90-year-old company built a monopoly on it. The AI boom just started. This stock is still flying under the radar.