Meta investment plans face scrutiny as capacity-contract claims point to higher 2027 capex, SemiAnalysis says

Meta Platforms’ next wave of spending on compute infrastructure is coming under fresh scrutiny, after a market report cited datacenter-capacity contracting and forecast a capex step-up in 2027, with SemiAnalysis pushing back on “neocloud” fears that demand projections are built on flawed assumptions.

The report, carried by Yahoo Finance, says Meta has contracted for more than 5 gigawatts of datacenter capacity across both cloud and colocation arrangements in the first half of 2026. In data-center terms, contracting for capacity at that scale generally reflects long-term commitments to power-hungry facilities used to run AI training and inference workloads, as well as core services.

Based on that contracting pace, SemiAnalysis argues that Meta’s capex trajectory is likely to be stronger in 2027. The Yahoo Finance piece frames the outlook as a response to worries in some parts of the industry that Meta’s compute needs, and therefore its spending trajectory, may have been overstated or misunderstood under “neocloud” narratives.

“Neocloud” is an industry term often used to describe a shift away from traditional cloud capacity procurement toward a model where hyperscalers and their partners increasingly tailor infrastructure to AI workloads, with tighter integration between hardware, power, networking, and software stacks. In the report’s framing, SemiAnalysis characterizes at least some of the neocloud-driven concerns as “erroneous,” though it does not, in the information available here, detail the underlying calculations.

Meta has publicly discussed its investments in AI infrastructure and the broader expansion of datacenter capacity in previous disclosures, but the new report’s specific capex timing and datacenter contracting totals are not presented as official company guidance in the material available for this story. As a result, readers are left to separate what Meta has confirmed from what third parties have inferred or compiled from contracting and capacity data.

The practical takeaway is that market participants may need to re-examine the shape of Meta’s infrastructure investment cycle, especially as AI compute demand becomes an increasingly central variable for operators of major datacenter portfolios. If the capacity-contracting figure and 2027 capex projection are directionally correct, it suggests Meta is positioning for continuing growth in compute supply at a time when data-center availability and power costs are major constraints across the sector.

Still, several key items remain unclear from the cited market report. It does not provide, in the available information, Meta’s own breakdown of expected 2027 capital spending by category, the remaining duration or pricing structure of the contracted capacity, or how much additional capacity Meta may seek beyond the first-half 2026 figure.

What to watch next is whether Meta’s upcoming disclosures, including any new datacenter or capital-spending guidance, align with the research firm’s timing. Investors and industry analysts will likely focus on changes in the company’s spending profile and on whether additional details confirm that a major portion of capacity is already committed far enough into the future to support a 2027 uplift.