In recent weeks, speculation has emerged regarding Meta’s potential transformation into a Neocloud provider, drawing intense reactions within the market. This speculation led to significant sell-offs of existing Neocloud entities like Coreweave and Nebius, triggering renewed discussions about possible “overcapacity” in the sector. However, analysts argue that these interpretations misrepresent the situation. They assert that Meta’s datacenter and compute procurement activities will not only accelerate but will also lead to substantial capital expenditures in the coming years.
In the first half of this fiscal year alone, Meta secured contracts for over 5 gigawatts (GW) of capacity across both cloud and colocation services, which does not factor in the company’s ramped-up self-build initiatives. The ongoing expansion of Meta’s infrastructural capabilities is underscored by its two largest datacenter campuses currently under construction, each representing 2.5 GW of additional capacity.
Contrary to alarming headlines which suggest that half of U.S. datacenters are delayed and only a mere 5 GW are presently being constructed, analysts point out that just two of Meta’s campuses account for a significant portion of this total capacity. These headlines are deemed inaccurate, reflecting an underappreciation of Meta’s aggressive development pace.
The abundance of compute capacity raises crucial questions: How will Meta deploy this newfound power? Will saturating the market with supply hinder their enterprise? Four major use-cases have emerged, deviating from traditional Neocloud strategies.
-
Frontier AI Models: Meta remains committed to advancing its frontier AI models, with a majority of its incremental capacity being directed to Meta Superintelligence Labs. Analysts are optimistic about Meta’s progress in this domain and plan to issue further insights on this unique data strategy’s performance against competitors like Anthropic and OpenAI.
-
Recommendation Systems (RecSys): There is an expectation that Meta can enhance its ads recommendation systems by over tenfold. Achieving this will necessitate a substantial amount of inference and training compute to refine its models. Furthermore, partnerships with firms like Anthropic may enable Meta to create a premium Sales & Marketing Software-as-a-Service product powered by frontier AI agents.
-
SpaceX-like Deals: Analysts predict that Meta aims to replicate the successful model established by SpaceX, which has carved a niche for high-demand, large-scale compute offerings at premium prices. Even modest capacity allocations could yield significant revenue returns—potentially exceeding $10 billion annually. A significant financial partnership with Anthropic may be on the horizon, marking the start of a lucrative revenue cycle.
-
Token-as-a-Service: Meta appears poised to launch a new token-as-a-service platform, potentially offering private instances of its AI models as a service. This move would allow for internal optimization as well as external partnerships, especially targeting larger enterprises seeking secure AI infrastructure.
The core of Meta’s compute strategy is deeply intertwined with its Superintelligence Labs. Still, the diversified potential for monetizing these resources presents an attractive backup plan for the company, should some initiatives underperform.
A detailed examination of Meta’s datacenter expansion throughout 2024 indicates that while about 10 GW of deals have been finalized recently, a significant portion now arises from third-party collaborations, enhancing yield prospects for partners like Coreweave and Nebius.
Moreover, the anticipated surge in other AI initiatives indicates multiple pathways for revenue generation, providing valuable monetization options even if the performance of Meta Superintelligence is lacking. The expected trajectory hints at a significant increase in compute supply—empowering Meta to explore varied lucrative avenues of engagement.
While the market has fixated on immediate headlines, the reality depicts a forward-looking Meta with grand ambitions across AI frameworks, prompting expectations of formidable growth in revenue and technological leadership.
Overall, the narrative surrounding Meta evolves amid wider discussions about the future of compute, capacity, and market dynamics, leaving industry observers eager to see how the unfolding strategy will shape its role in the AI ecosystem.
