Meta launches enterprise AI cloud service; proprietary Iris chip starts September production
Meta has launched Meta Compute to offer AI infrastructure and foundation models as a service, aiming to compete with major cloud providers. Additionally, the company confirmed that production of its proprietary Iris AI chip is scheduled to begin in September 2026 to optimize internal training and inference efficiency.
Key Takeaways
- Meta Compute enters the hyperscaler market, offering raw GPU access and hosted foundation models to compete with AWS, Azure, and Google Cloud.
- The proprietary Iris AI chip, developed with Broadcom and TSMC, starts production in September to reduce reliance on NVIDIA and AMD silicon.
- Wolfe Research estimates every gigawatt of AI capacity successfully monetized through Meta Compute could boost earnings per share by 20%.
- Internal data center capacity is projected to scale from approximately 7 gigawatts in 2026 to 14 gigawatts by the end of 2027.
- Meta guided Q2 revenue to $58–61 billion ahead of its July 29 earnings report, where analysts expect a critical update on peak capital expenditure.
Why It Matters
Meta is evolving from a pure-play advertising giant into a vertically integrated infrastructure provider. By commercializing excess data center capacity and deploying custom silicon, Meta can potentially lower its infrastructure unit costs while opening a secondary revenue stream outside of volatile ad markets. This pivot forces traditional cloud incumbents to defend existing enterprise margins against a competitor that has already amortized its buildout costs for internal use. For the broader ecosystem, it signals that large-scale inference and training are becoming commodity utilities. Watch the July 29 capital expenditure guidance; any increase beyond the current $125–$145 billion range will test investor patience regarding the ROI timeline for these 14-gigawatt ambitions.
Additional Context
The launch of Meta Compute follows a broader strategic reorganization that prioritized industrial-scale infrastructure. Per SemiAnalysis in July 2026, Meta is currently building five individual one-gigawatt 'Titan' data center clusters, including the Prometheus site in Ohio, to maintain a compute advantage over OpenAI and Anthropic. This infrastructure blitz is backed by an unprecedented recruitment drive; reports indicate Meta has spent billions to poach researchers from rivals with compensation packages occasionally exceeding $1 billion. Additionally, the company significantly deepened its data pipeline through a $14.3 billion deal in 2025 to acquire a near-majority stake in Scale AI, securing high-quality reinforcement learning feedback cycles. While infrastructure momentum is high, Meta's consumer-facing AI products have faced immediate friction. Just days after its July 2026 debut, Meta withdrew a key feature from its Muse Image tool following intense privacy backlash. Per The Guardian and Reuters, the feature had automatically opted in public Instagram accounts to allow third-party users to generate synthetic images of them using @mentions. SAG-AFTRA and other industry groups labeled the default opt-in an 'utter miscalculation' of public sentiment regarding digital replicas. Although Meta removed the feature, the company continues to rollout secondary Muse-powered effects across its apps, showing that regulatory and privacy hurdles remain the primary drag on its otherwise bullish technical roadmap.
Read full article at tradingkey.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source