AI chip stocks rebound on multiyear deal flow as analysts lift forecasts

The sell-off in AI chip stocks looked alarming while it lasted. The recovery, it turns out, was waiting on a fairly simple piece of evidence: customers were signing contracts that stretched years into the future, and that is not how companies behave when they expect demand to collapse.

The contract evidence that moved markets

Micron anchored the optimism with a striking figure: $22 billion in multiyear customer commitments for memory chips, secured as of mid-September 2026.

Those are not handshake deals. Many are structured as take-or-pay agreements, meaning customers owe payment whether or not they ultimately take delivery.

Broadcom added its own data point, announcing an extended partnership with Meta running through 2029. The company’s CEO described current AI demand as exceptionally strong, projecting AI semiconductor revenue of $115 billion for fiscal 2027 and $230 billion for fiscal 2028.

Amazon reinforced that picture by raising its full-year capital expenditure forecast to $220 billion.

Piper Sandler moved quickly on the data, initiating Overweight ratings on Nvidia, AMD, and Broadcom with price targets of $300, $600, and $460 respectively. Bank of America went further, raising its U.S. semiconductor market growth forecast from 14% annually to 18% annually through 2030.

Why multiyear deals change the calculus

Analysts have flagged that chip supply tightness is expected to persist into 2027, which makes early commitments a competitive necessity rather than a speculative bet.

The AMD-Anthropic partnership, in which AMD secured a deal to supply Instinct GPUs to the AI startup, fits this pattern. So does Nvidia’s expanded optical manufacturing work with Corning.

Bank of America’s revised forecast projects the U.S. semiconductor market reaching roughly $3.2 trillion by 2030. Piper Sandler’s broader view puts the global AI compute market at $2.2 trillion by the same year.

What the rebound means for the broader tech landscape

Micron’s $22 billion commitment backlog provides a degree of revenue visibility that a chip company without such agreements simply cannot claim. Broadcom’s Meta partnership through 2029 creates a multi-year floor under its AI revenue line.

The risk to this thesis is concentration. When a handful of hyperscalers account for the majority of AI infrastructure spending, their individual capital allocation decisions can move entire markets. The multiyear contracts provide some insulation, but they do not eliminate that dependency entirely.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.