OpenAI Builds AI Chips With Broadcom, Nvidia Falls 3.25%

OpenAI is developing its own artificial intelligence chips in partnership with semiconductor designer Broadcom, according to a report published June 26. The strategic initiative to reduce reliance on external suppliers contributed to a sell-off in Nvidia shares, which fell 3.25% to trade at $192.53 as of 07:48 UTC today. The stock traded within a range of $191.22 to $195.55 amid the news-driven volatility.

Context — why this matters now

The move follows a pattern of major technology firms designing custom silicon to optimize performance and control costs for massive AI workloads. Google launched its Tensor Processing Unit (TPU) in 2016, and Amazon Web Services introduced its Trainium and Inferentia chips in 2020. Microsoft unveiled its Maia AI accelerator and Cobalt CPU in November 2025. This vertical integration trend accelerates as the computational demands of large language models double every few months, making hardware efficiency a critical competitive advantage.

The current macro backdrop features intense capital expenditure from cloud providers on AI infrastructure. This has driven record revenue for chip suppliers like Nvidia, which reported $26.7 billion in data center sales last quarter. However, the soaring cost of procuring thousands of high-end GPUs is compressing margins for AI service providers. Designing custom chips represents a long-term strategy to alleviate these cost pressures and secure supply chain independence.

Data — what the numbers show

Nvidia's stock declined 3.25% to $192.53 on the session, underperforming the broader technology sector. The stock's intraday low of $191.22 approached a key technical level last tested in May. Nvidia's market capitalization dropped approximately $80 billion in the session based on its outstanding shares.

This pullback contrasts with the PHLX Semiconductor Index (SOX), which was down a more moderate 1.8% over the same period. Broadcom shares showed relative resilience, declining only 1.2%, as the company stands to gain design wins and associated engineering revenue from the partnership. The collaboration represents a significant, though undisclosed, financial commitment from OpenAI for the development phase.

Analysis — what it means for markets / sectors / tickers

The development signals a potential long-term risk to Nvidia's dominance in the AI training market, though any material impact remains years away. Custom chip initiatives require extensive design, validation, and manufacturing cycles, often taking two to three years to reach production scale. In the immediate term, Nvidia continues to benefit from its entrenched software ecosystem, CUDA, which locks in developers and creates high switching costs for alternative hardware.

Second-order effects include potential benefits for semiconductor intellectual property firms like Arm Holdings and Synopsys, which provide essential design tools and architectures. Taiwan Semiconductor Manufacturing Company also stands to gain from manufacturing these new chips, solidifying its role as the world's advanced foundry. Trading flow data indicated elevated put option activity on Nvidia, with volume doubling the 20-day average, suggesting some investors are hedging against further downside.

Outlook — what to watch next

Key catalysts include Nvidia's next earnings report on August 21, where management will likely address the competitive landscape and any potential customer concentration risks. Investors should monitor comments on the durability of its software moat and the adoption rate of its next-generation Blackwell architecture chips.

Technical levels to watch for NVDA include the 50-day moving average near $188, which has provided support during previous pullbacks. A break below this level could signal a deeper correction. For Broadcom, investors will focus on its July 11 earnings call for any updates on its custom silicon design business and revenue projections from AI-related projects.

Frequently Asked Questions

What does OpenAI building its own chips mean for Nvidia's business?

The initiative poses a long-term competitive threat to Nvidia's data center revenue, which reached $26.7 billion last quarter. However, the development cycle for custom AI chips typically spans multiple years, and Nvidia maintains a significant advantage through its proprietary CUDA software platform. Most AI developers are trained on and optimized for Nvidia's ecosystem, creating substantial switching costs that protect its market position.

How does this development compare to previous tech companies designing custom chips?

This follows established precedents from cloud infrastructure leaders. Google has deployed its TPUs since 2016, and Amazon's Graviton processors now power most of its EC2 instances. Microsoft's Maia AI accelerator is scheduled for deployment in 2026. The key difference is OpenAI's position as an AI application company rather than a cloud infrastructure provider, suggesting even AI software firms now view vertical integration as strategically necessary.

Will other AI companies follow OpenAI in developing custom chips?

Leading AI labs with sufficient capital resources may pursue similar strategies to control costs and optimize performance. Anthropic and Mistral AI would be the most likely candidates given their scale and funding. However, the immense R&D expenditure required—often exceeding $1 billion per design—creates a significant barrier to entry, limiting this option to only the best-funded private companies or those backed by major cloud providers.

Bottom Line

Nvidia faces nascent competition from vertical integration, but its software ecosystem remains a durable advantage.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.