Custom Chips and a $7B Dilution Reset the AI Hardware Trade

The setup — The hardware trade is facing a sharp reality check as the sheer cost of artificial intelligence starts to bite. Model builders are designing their own silicon to escape single-supplier pricing, while legacy chipmakers are paying steep equity premiums to buy their way into new verticals. It all comes to a head in a severe repricing for one analog powerhouse trying to pivot its way to the edge.

What's moving

OpenAI is stepping out of the single-supplier ecosystem, partnering with Broadcom $AVGO to build a custom inference chip dubbed Jalapeño (per TechCrunch). The move positions the leading AI lab alongside Google and Apple in a growing cohort of tech giants designing custom silicon to break their total reliance on Nvidia $NVDA hardware. The shift signals that the next phase of the infrastructure buildout will prioritize efficiency and unit economics over raw compute speed.

Software multiples caught a bid as the immediate disruption threat from foundation models began to cool, lifting shares of ServiceNow $NOW and Salesforce $CRM (per MarketWatch) into the green. Enterprise buyers are shifting their focus from broad AI experimentation to budget efficiency and demonstrable returns on investment, giving traditional cloud software vendors breathing room to defend their subscription models without being entirely displaced by generative tools (per CNBC).

The Trump administration walked back a strict export control directive, granting more than 100 select organizations and government agencies access to Anthropic's Mythos 5 model (per TechCrunch). The model had been previously taken offline for select enterprise use citing national security authorities. As Sudden Export Rules Reveal the Fragility of AI Dominance, the back-and-forth negotiation underscores the rising political stakes of frontier models and the unpredictable regulatory environment for hyperscalers trying to commercialize them (per The Verge).

Featured: ON Semiconductor Corporation (ON)

The move
ON Semiconductor $ON erased nearly a quarter of its value, falling 23.66% from a prior close of $118.74 down to $90.65. The aggressive selloff broke the stock decisively below its key moving averages, though the equity remains up roughly 67% on the year.

What drove it
The company signed a definitive agreement to acquire edge-AI and connectivity designer Synaptics $SYNA in a $7 billion all-stock transaction (per Barchart). Existing shareholders will foot the bill, with Synaptics investors receiving 1.35 ON shares for every piece of paper they currently hold. Management pitched the deal as a necessary expansion, with the CEO stating the pivot into physical AI components grows their total addressable market by an additional $30 billion (per CNBC). The market took one look at the dilution and the strategic rationale, and aggressively rejected the premise.

The bigger picture
This is a story about margin profiles and the dangers of style drift. ON Semiconductor built its premium valuation — trading near 38 times forward earnings — by dominating high-margin, sticky end markets like automotive power switching and industrial automation. They sell components that are difficult to design and rarely replaced once integrated into a vehicle or factory floor.

Synaptics, by contrast, operates in the brutal, highly cyclical world of consumer electronics. The market is violently punishing ON Semiconductor for using expensive equity to buy its way into a lower-quality revenue stream under the guise of an AI pivot. Investors are perfectly willing to pay a premium multiple for industrial consistency, but they will actively dump shares when a company dilutes that consistency to chase consumer edge devices. It signals to the street that the company may fear growth is slowing in its core automotive verticals, prompting a desperate, expensive reach into adjacent markets just as consumer hardware faces mounting supply chain costs.

Across the tape

Apple $AAPL raised prices on Macs and iPads in the middle of Amazon Prime Day, directly passing the rising cost of memory chips onto consumers as the AI-driven component shortage begins to squeeze traditional electronics (per The Verge).

Asian technology stocks led a broad global selloff amid mounting anxiety over the sheer capital expenditure required to build out data centers for artificial intelligence (per CNBC).

Prediction markets have dramatically repriced the timeline for an OpenAI public debut, pushing expectations into early next year or as late as June 2027, despite the company confidentially filing a prospectus with the SEC earlier this month (per CNBC).

Bond yields continue to drift lower as the Federal Reserve’s new leadership adopts an aggressive rhetorical stance on inflation, helping to coax Treasury rates down despite recent data prints (per MarketWatch).

What to watch

• ON Semiconductor's automotive margins: Watch their upcoming quarterly print for weakness in the core automotive segment. If base margins are contracting, the street will view the Synaptics deal as a defensive scramble rather than an offensive expansion.
• Consumer electronics pricing data: With Apple raising prices, watch whether secondary laptop and smartphone makers follow suit in the coming weeks to offset their own memory chip acquisition costs.
• Enterprise software retention rates: As companies tighten their generative AI budgets, watch upcoming software earnings to see if cloud infrastructure platforms can maintain their historical subscription backlogs.
• Nasdaq fast-track inclusion updates: Keep an eye on the exchange's new regulatory framework, which is reportedly paving the way for immediate inclusion of major private entities upon listing (per CNBC).