Is AI Compute Overcapacity a False Narrative? Jefferies Upgrades Meta: AI Investment Just Beginning, Target Price Raised to $825

Investment bank Jefferies has released a new research report stating that recent market concerns over a potential slowdown in Meta Platforms' (META-US) capital expenditures have been overinterpreted. The firm argues that Meta's decision to sell some of its idle AI computing power does not indicate oversupply or reduced investment, but rather reflects a strategic effort to improve infrastructure utilization efficiency and expand enterprise cloud service capabilities. In the long term, this could even replicate the successful model of Amazon's (AMZN-US) AWS, creating a second growth curve for the company.

The report, published by Jefferies senior internet analyst Brent Thill on July 8, argues that Meta's AI infrastructure build-out is "far from over," and that current market skepticism around AI capital spending is misplaced. The company is gradually transforming its massive AI infrastructure investments into new businesses with strategic value and commercial returns.

Prior to the report's release, Meta's stock price had experienced significant volatility. On July 1, shares surged nearly 9% after media reports suggested Meta was considering leasing out part of its AI compute capacity. However, the next day, investors worried about excessive AI investment and slowing demand, causing the stock to drop nearly 4%, dragging down AI infrastructure-related stocks. CoreWeave's shares fell 13% in a single day, while Nebius dropped 15%.

Nevertheless, Jefferies contends that recent statements by Meta's infrastructure lead at the Scale AI Conference indicate that the company's AI development remains in its early stages. The company's first large-scale, multi-gigawatt AI cluster, "Prometheus," officially went live in July and is now being used for training large AI models, demonstrating Meta's ongoing expansion of AI computing power.

The report notes that the real bottleneck in the AI industry has long moved beyond GPUs and is now expanding to CPUs, memory, storage, and overall data center infrastructure, with supply still far below market demand.

Citing industry surveys, Jefferies states that the AI compute capacity of the three major cloud providers—Alphabet (GOOGL-US), Amazon (AMZN-US), and Microsoft (MSFT-US)—is almost entirely sold out for the next six months. Not only is there no price competition, but rental prices for compute power are actually rising, indicating that the feared "compute overcapacity" does not exist in the market.

The report estimates that Meta's current AI infrastructure utilization rate is around 65%, leaving some idle capacity available for external customers. Jefferies believes this demonstrates Meta's flexibility to develop new businesses, rather than signaling overinvestment.

Analyst Brent Thill points out that the combined backlog of AI compute demand across the three major cloud platforms totals approximately $2 trillion, representing a massive market opportunity for any new entrant with sufficient capital and technical capability.

Jefferies forecasts that by 2026, the four hyperscalers—Microsoft (MSFT-US), Amazon (AMZN-US), Alphabet (GOOGL-US), and Meta Platforms (META-US)—will spend between $620 billion and $770 billion annually on capital expenditures, representing a year-on-year increase of 65% to 74%, reflecting the ongoing acceleration of the AI infrastructure race.

The report compares Meta's strategy to AWS's early development model. AWS initially built computing resources to support Amazon's e-commerce peak demand and later leased out idle capacity, eventually evolving into the world's largest cloud service platform.

Jefferies believes Meta can similarly transform its AI compute power from a cost center into a revenue source, improving data center utilization and return on invested capital (ROIC), generating more free cash flow, and reinvesting it into even larger-scale AI infrastructure, creating a virtuous cycle rather than reducing capital spending.

In terms of valuation, Jefferies notes that Meta's current forward P/E ratio of around 18x remains attractive compared to other large tech stocks. The report forecasts the company's earnings per share (EPS) could reach $35 and maintains an $825 price target, implying approximately 41% upside from the current stock price.

Jefferies also highlights that even if AI and cloud businesses have not yet significantly contributed to profits in the short term, Meta's core advertising business remains robust. The company reported first-quarter 2026 revenue of $56.3 billion, up 33% year-on-year, continuing to provide strong cash flow to support AI investments.

Additionally, Jefferies' industry survey shows that as high as 95% of surveyed enterprises expect their cloud budgets to increase in 2026, and global cloud spending growth is projected to rise from 9.6% in 2025 to 10.1%, reflecting that AI-driven data center demand continues to grow rapidly.

However, the report warns investors to be mindful of execution risks. The analyst admits that Meta's entry into the enterprise cloud services market is "very late." With Amazon's AWS already capturing nearly 50% of the global cloud market, Meta will need to build enterprise sales teams, customer service systems, and enterprise security capabilities, likely starting with small and medium-sized business customers before gradually challenging large enterprises.

Jefferies also notes that the four major cloud providers have already issued approximately $144 billion in bonds this year, exceeding the $83 billion issued in all of 2025, indicating that AI infrastructure development is increasingly reliant on debt financing. If future AI investment returns fall short of market expectations, the hyperscale capex cycle could face correction pressure.

Moreover, the report mentions that Meta's current AI model capabilities still lag behind the most advanced models. If it fails to continuously close the gap with competitors, this could impact the competitiveness of its future enterprise cloud services.

On the other hand, Alphabet's Google has recently restricted Meta's access to Gemini AI model resources due to its own tight compute supply, highlighting that AI compute has become a critical strategic resource among tech giants. Jefferies believes that once Meta formally enters the enterprise cloud market, it will face more direct competition with Amazon, Microsoft, and Alphabet, potentially leading to price competition or exclusive partnerships in the future.

Overall, Jefferies argues that the market has misinterpreted Meta's external provision of AI compute as a sign of cooling AI investment. In reality, the company is leveraging higher utilization and enterprise cloud services to build a new business model capable of supporting even larger AI investments. Against the backdrop of severe global AI compute supply shortages, Meta's long-term AI strategy and enterprise cloud business still hold substantial growth potential.