$10B OpenAI–Broadcom chip deal aims to power GPT-5, cut Nvidia dependence

Close-up dual editorial portraits of Sam Altman and Hock Tan facing slightly toward each other above a luminous silicon wafer bridge with intricate glowing circuitry, bright high-key lighting, warm amber and cool electric blue palette, shallow depth of field, neutral expressions, crisp studio clarity.

OpenAI is reportedly preparing to mass-produce its own AI chips in 2026 in a partnership with Broadcom, a move intended to meet rising compute needs and reduce reliance on Nvidia. The chips are expected to be shipped next year, according to multiple sources cited in the report.

Broadcom partnership and shipping timeline

According to Verdict, the chip effort is being developed with Broadcom, and insiders say OpenAI plans to use the hardware internally rather than offering it to external clients. The partnership began in 2024, though the timeline for mass production had not been previously clear.

Verdict reports that Broadcom CEO Hock Tan told analysts the company had secured a fourth customer for its custom AI chip division, without naming the client. While Broadcom does not disclose customer identities, sources confirmed that OpenAI is the new customer, the publication said. Both companies declined to comment, according to the report.

Tan said the agreement has improved Broadcom’s growth outlook by creating “immediate and fairly substantial demand,” with plans to begin shipping chips for this customer “pretty strongly” from next year. Broadcom has previously collaborated with Google to develop custom “TPU” AI chips.

Context: compute needs and industry moves

Internal use to support ChatGPT and model training

Verdict notes that OpenAI intends to deploy the chips internally to support workloads including products such as ChatGPT and the training and execution of AI models. The approach would align OpenAI with firms like Google, Amazon, and Meta, which have developed their own specialized chips for AI.

OpenAI CEO Sam Altman has emphasized the need for increased computing power as usage grows. In August 2025, Altman indicated the company is prioritizing computing resources “in light of the increased demand from [OpenAI’s latest model] GPT-5” and aims to double its computing fleet “over the next five months,” according to the report.

The report adds that Broadcom’s CEO recently referenced a new undisclosed customer that committed to $10bn in orders. While the company did not name the buyer, sources identified OpenAI as the fourth customer for Broadcom’s custom AI chip division. Verdict states that both OpenAI and Broadcom chose not to comment on the matter.

Total
0
Shares
Pridaj komentár

Vaša e-mailová adresa nebude zverejnená. Vyžadované polia sú označené *

Previous Post
A bright, high-contrast world map grid with the outline of China glowing red-orange, a large polished steel padlock hovering above it, blue-white network lines rerouting around the region, soft haze and rim light, clean minimal background, close to medium-wide top-down view.

U.S. AI Giant Anthropic Expands Restrictions on China-Linked Users Amid Geopolitical Tensions

Next Post
A luminous gold custom silicon chip hovering at center like a magnet, pulling in a swirl of smaller dark chiplets and glowing data threads from the edges, dramatic warm gold versus deep indigo-blue background, crisp high-key lighting, clear central composition, no text or logos

Mystery $10B AI chip deal sparks OpenAI speculation; Nvidia, AMD slide

Related Posts