OpenAI is pushing beyond chatbots and coding assistants into specialized industries such as chip design, life sciences and financial services, while its coding platform Codex has reached 25 million users, according to OpenAI CFO Sarah Friar.
The announcement highlights a broader shift in OpenAI’s strategy: instead of selling AI primarily as a general-purpose assistant, the company increasingly wants its models to perform highly specialized work for businesses.
One of the strongest examples is semiconductor design. OpenAI says it used its own AI models during development of its Jalapeño chip, helping take the design to the “tape-out” stage in nine months.
OpenAI Is Taking AI Into Chip Design
OpenAI is positioning its models as tools that can assist engineers with specialized technical workflows.
CFO Sarah Friar said the company is focusing on industries where AI can produce measurable business outcomes, including semiconductor design, life sciences and financial services.
The chip-design push is particularly significant because semiconductor development is traditionally a complex, highly specialized process involving architecture, verification, optimization, physical design and manufacturing preparation.
OpenAI’s own experience developing Jalapeño is being used as evidence that its models can contribute to this type of technical work.
However, the claim should be understood carefully: OpenAI is not simply announcing that an AI model independently designed an entire production chip. Friar said OpenAI used its models in developing Jalapeño and that the chip reached tape-out within nine months.
What Does “Tape-Out” Mean?
In semiconductor development, tape-out is a major milestone.
It means the chip’s design has been finalized sufficiently to be sent to a manufacturing facility for fabrication.
Reaching tape-out does not mean that manufactured chips are already shipping to customers. After tape-out, the design still has to go through fabrication, packaging, testing and validation.
That makes the nine-month timeline notable because OpenAI is effectively presenting its own chip project as a demonstration of how AI could accelerate parts of the semiconductor development process.
The company is building Jalapeño as an AI inference accelerator, with the project involving partners including Broadcom.
Codex Reaches 25 Million Users
While OpenAI is moving into hardware and specialized enterprise applications, its software business is also expanding rapidly.
OpenAI’s Codex coding platform has reached 25 million users, according to Friar.
Codex is designed to help developers work with software using AI, including generating, modifying, analyzing and working through coding tasks.
The milestone shows how coding has become one of the largest practical applications for frontier AI.
It also strengthens OpenAI’s argument that its models can move beyond conversational assistance into workflows where users expect the AI to actually perform technical work.
Why Codex Matters to OpenAI’s Chip Strategy
The connection between Codex and chip design is more important than it may initially appear.
Modern semiconductor development depends heavily on software, automation, simulation and engineering tools.
AI systems that can reason through code, inspect complex technical specifications and automate portions of engineering workflows could potentially reduce the amount of manual work required during chip development.
OpenAI’s experience with Jalapeño provides an internal test case for that strategy.
If the company can demonstrate that its own models accelerate chip development, it could potentially offer similar AI capabilities to semiconductor companies, engineering teams and other highly technical enterprises.
OpenAI Is Targeting Specialized Enterprise Work
Friar said OpenAI is increasingly focused on specialized industries rather than treating every customer as a generic AI user.
Chip design is one example. Life sciences and financial services are two other areas the company is targeting.
This reflects a broader trend across the AI industry.
Businesses are increasingly asking whether AI can produce measurable improvements in productivity, revenue or operating costs rather than simply providing access to a powerful chatbot.
OpenAI is responding by exploring outcome-based pricing, where customers could potentially pay according to the business results delivered rather than simply the amount of AI they consume.
OpenAI Is Also Cutting AI Costs
The expansion comes as OpenAI faces intense competition from open-weight AI models.
Friar said OpenAI recently reduced the price of its lower-cost Luna model by 80%, which was followed by approximately a 10-fold increase in usage.
She also argued that deploying Luna through cloud infrastructure can be cheaper than running some competing Chinese open-weight models through cloud providers.
The pricing strategy suggests OpenAI is trying to defend its position from both ends of the market: increasingly powerful frontier models at the high end and cheaper open-weight alternatives at the lower end.
Enterprise AI Is Becoming a Bigger Part of OpenAI
OpenAI’s enterprise business is also growing rapidly.
Friar said enterprise revenue increased 32% from June to July, compared with approximately 20% growth in overall annualized revenue during the same period.
She also said enterprise and consumer revenue had reached roughly an even split by the middle of the year, ahead of OpenAI’s previous target of achieving that balance by the end of the year.
That shift matters because specialized AI products can potentially generate larger and more durable business relationships than consumer chatbot subscriptions alone.
OpenAI’s Bigger Strategy Is Coming Into Focus
The combination of Codex’s 25 million users, AI-assisted chip development and specialized enterprise offerings points toward a broader strategy.
OpenAI increasingly wants its models to become infrastructure for technical work.
That could include:
- Software engineering
- Semiconductor design
- Scientific research
- Financial analysis
- Business automation
- Enterprise workflows
- AI-powered engineering
The company’s own Jalapeño project is particularly important because it provides an example of OpenAI using its AI internally before attempting to sell similar capabilities externally.
What This Means for the AI and Chip Industries
If AI can meaningfully accelerate chip development, the impact could extend far beyond OpenAI.
Chip companies could use AI to explore more architectures, automate repetitive engineering tasks and identify design problems earlier.
For AI companies, the benefit could be even larger.
The AI industry currently depends heavily on increasingly sophisticated and expensive semiconductor infrastructure. Improving the speed at which custom accelerators are designed could eventually give AI companies greater control over their hardware roadmaps.
OpenAI is therefore trying to compete not only at the model layer, but increasingly across the software, enterprise and hardware stack.
Read More:- Meta Launches Muse AI Agent That Can Email, Shop and Make Payments for Users
The Bottom Line
OpenAI’s latest move shows how quickly the role of AI is expanding.
Codex reaching 25 million users demonstrates the scale of AI-assisted software development, while the Jalapeño project offers an early example of OpenAI using its own models in semiconductor development.
The bigger story is not simply that OpenAI is building a chip.
It is that OpenAI increasingly sees its models as tools capable of performing specialized technical work across entire industries.
If that strategy succeeds, the next phase of the AI race may be less about who has the best chatbot and more about which AI company can automate the largest amount of real-world engineering and enterprise work.
