Chip stocks jump after OpenAI-Broadcom chip deal

Chip stocks rose on Monday after OpenAI and Broadcom announced a deal to develop custom AI chips and concerns over US-China tensions eased.

Broadcom shares climbed 10% following the announcement.

The VanEck Semiconductor ETF gained nearly 4%, with Nvidia and AMD each up over 3%. Taiwan Semiconductor and On Semiconductor jumped 8% and 9%, while Micron Technology rose more than 6%.

OpenAI, led by Sam Altman, recently signed chip supply agreements with Nvidia and AMD.

Market sentiment improved after President Donald Trump said on Truth Social that issues with China “will all be fine,” easing fears after his earlier tariff threats.

Major tech firms, which lost US$770 billion in market value during October 10’s selloff, saw Alphabet and Tesla recover those losses on October 13.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

OpenAI’s chip plan still relies on familiar suppliers

The 10-gigawatt (data center power capacity) Broadcom deal starts deployments in late 2026 with a full rollout by the end of 2029 1, so near-term compute needs stay on Nvidia and AMD systems already under contract.
Total hardware commitments reach about 33 gigawatts across Nvidia and Broadcom 2. Oracle (a cloud infrastructure provider) and AMD are included as well 2. Current operations sit just over 2 gigawatts, so these deals plan capacity over time rather than an immediate shift.
Broadcom is building OpenAI accelerators using its Ethernet networking plus hardware IP (reusable intellectual property blocks) 3. Foundry partners (contract chip manufacturers) and packaging details are undisclosed 3, so the scope of any shift away from Taiwan Semiconductor advanced packaging remains unclear.

Advanced packaging shifts open doors for equipment suppliers

Taiwan Semiconductor CoWoS (Chip-on-Wafer-on-Substrate) places High Bandwidth Memory (HBM) near processors, runs near 60% utilization despite AI demand, and could reach about 100,000 units per month by late 2026 4.
Vendors tied to advanced packaging may benefit as a second wave of custom chips lands after 2026, especially those aligned with AP8 plus upcoming AP9 and AP10, designations for Taiwan Semiconductor advanced packaging sites 4.
Firms selling chiplet interconnects (die-to-die links that let multiple small chips act as one) plus related system parts could gain share 3. Custom accelerator buyers want parity with Nvidia CUDA, while hyperscalers (the largest cloud providers) and AI labs pursue differentiation without large in-house design teams 3.