Report and timing:
Simply Wall St (Feb 16, 2026) flagged Meta's plan for a $10B data-center build in Indiana and a tilt toward more closed-source AI, framing it as a test of the stock's valuation. Meta officially broke ground on a 1-gigawatt (GW) campus in Lebanon, Indiana on Feb 11, 2026, calling it one of its largest infrastructure bets to power AI and core apps. Read the market report on Yahoo Finance (UK).
What's actually being built:
This is full-stack compute, not just a pile of GPUs. Meta says the Lebanon campus targets about 1 GW capacity and includes buildings, servers, networking, power distribution and advanced cooling. The plan also includes over $120 million in local water and utility upgrades, roughly 4,000 peak construction jobs and about 300 operations roles. This is capital expenditure (capex) for land and shells plus the hardware to train and serve AI models. See Meta's announcement at Meta Newsroom.
Open vs closed, in plain English:
"Open-source" in AI usually means releasing model weights so anyone can run or fine-tune the model. "Closed-source" means keeping those weights private and providing access through APIs. Mark Zuckerberg has signaled Meta will not open-source all "superintelligence" models, so expect a mix: open releases like Llama plus proprietary frontier systems. That approach can improve control and monetization, but it limits interoperability and third-party fine-tuning. For background, see TechCrunch.
Valuation context, not vibes:
Per Simply Wall St, Meta traded around $640-$650 in mid-February, roughly 25-32% below the $860.08 analyst target midpoint and about 40% below its discounted cash flow (DCF) fair value estimate near $1,080. Meta's price-to-earnings (P/E) ratio is about 26.8x versus 11.2x for the U.S. Interactive Media & Services industry. Three-year returns are about 276%, with roughly 3% 30-day momentum. Translation: big infrastructure spending, but in the Simply Wall St model the stock still "screens cheap." See their valuation page at Simply Wall St.
Regulatory and partner risk:
Tightening model access could invite more scrutiny from regulators and ecosystem partners on safety, access, and competition. That may slow the rollout of new AI features and complicate partnerships. This risk sits alongside an expanding infrastructure footprint. More context at TechCrunch.
Why founders should care:
Sales reality: A $10B campus raises the bar for AI infrastructure vendors. Expect tougher requests for proposals (RFPs), longer sales cycles, and price pressure as hyperscalers chase 1-GW-class capacity. See Meta's project details at Meta Newsroom.
Platform strategy: If cutting-edge models go closed, plan for API lock-in. Keep fallback paths across providers and track cost-per-inference, latency service-level objectives (SLOs) and data egress terms on a monthly cadence. Guidance at TechCrunch.
Talent market: builds like this absorb machine-learning/operations (ML/ops) specialists. Calibrate hiring with remote benches, equity incentives, and partnerships with managed model providers. See hiring and project notes at Meta Newsroom.
Local context: This is Indiana site No. 2 after Jeffersonville’s $800M build. It shows compute is moving where land and power are available, not just the coasts. Local reporting: AP News.
One more thing:
All of this is based on reported plans and market analyses. Meta has not published a phase-by-phase capex breakdown or a final openness policy for future models, and timelines can shift. Some reports point to initial operations in 2027-2028. For construction reporting, see Cinco Días / El País.
Get daily insider tech news delivered to your inbox every weekday morning.
