Intel Expands Secondary Offering to $20 Billion, Eyes AI Demand Surpassing $100 Billion

Intel set the price for its primary offering of common shares at $95 per share on the evening of Monday, August 10, raising the size to $20 billion, $5 billion more than the company had proposed in the morning. CEO Lip-Bu Tan took this opportunity to raise cash at a pace that the American chipmaker hasn't attempted since its listing in 1971.

The book of orders exceeded $100 billion, Bloomberg reported, five times the offered volume. Coordinators received the option to purchase up to an additional $2.25 billion within 30 days, which could elevate the fundraising to over $22 billion. Intel projects net proceeds of approximately $19.7 billion after commissions, according to the market announcement, and the resulting dilution for current shareholders will be about 3%.

A Shift in Issuance Tone

Established chipmakers tend to prefer debt over equity. Issuing shares dilutes ownership, and Intel is entering 2025 under pressure to cut costs, not to bolster equity capital. Tan took over earlier this year with a mandate to straighten the finances and has attracted, over the past months, capital from the U.S. government and a cross-investment from Nvidia, an unusual move among direct competitors.

What has changed is the demand for wafers for model inference. Revenue from the data center and AI segment grew by 59% year-on-year in the previous quarter, and areas related to artificial intelligence advanced over 70% in aggregate. The stock surged 164% this year, reaching $101.65 last week, and Tan decided to monetize part of this re-rating before the market re-evaluates the sector.

The funds will be utilized for what the company describes as "general corporate purposes," with a stated focus on new applied AI markets and custom silicon. Intel has already raised its 2026 capex guidance to above $20 billion in July and signaled to investors that 2027 will require an even larger investment. Without this fundraising, a significant portion of this would need to be financed with debt, which is expensive in a scenario of U.S. interest rates above 4%.

Asian Competitors Watch and Pay the Cost

In South Korea, the government of Lee Jae-myung announced on Monday a fund of 5 trillion won, approximately $3.5 billion, to accelerate local semiconductor hubs. This is accompanied by another round of 5 trillion won in commercial financing for suppliers of Samsung and SK Hynix. It is a response to the megaproject of $576 billion announced in June by the two Koreans alongside regional governments.

In Taiwan, TSMC continues with its roadmap for N2 node expansion and has committed to $165 billion in factories in the U.S. under the Chips Act umbrella. The competition with Intel in contract manufacturing has become more than just rhetoric. Among the few names operating nodes below 3nm at volume, Intel Foundry is the only one whose capacity was constrained due to lack of capital before this offering.

What Changes for the Corporate Buyer

For CIOs, the implications are twofold. On one hand, Intel Foundry now has resources to challenge Nvidia and AMD with vertically integrated AI platforms, opening up a new entrant in multi-year infrastructure contracts. On the other hand, the job market for silicon engineers in hubs like Guadalajara and Penang is likely to tighten, with Samsung and TSMC driving salaries up in response.

In Brazil, the effects are felt through two concrete channels. Local consumers of GPU capacity, such as Petrobras and large banks, currently contract capacity that passes through American hyperscalers. A more aggressive Intel on price pressures Nvidia and marginally lowers cloud training costs in local markets. On the industrial side, PC manufacturers in Sorocaba and Manaus depend on a stable supply of Intel chips, and increased global capacity reduces the risk of a new allocation crisis like the one that disrupted shipments in 2021.

The offering arrives at a moment when Tan needs to convert financial demand into operational results. The timeframe the market has given for this conversion is short. The $100 billion in orders came under the premise that AI continues to consume silicon at a geometrically increasing rate. If this slows down by 2027, the dilution will have been high.