"Intel Inside." At one time, there was nothing better. It was the seal of innovation and computing power. For a long time, Intel reigned supreme in the processor market, powering PCs worldwide. In the collective imagination, AMD played second fiddle, condemned to low-end machines and broke geeks. Intel was innovation, stability, performance, the standard. But those days are gone. Today, the Santa Clara giant is dragging its carcass around like a fallen champion.

Still there, but tired. Still powerful, but disoriented. It remains a pillar of the industry, a builder, certainly, but a builder in the midst of an identity crisis. For a long time, Intel wanted to do everything: design its chips (CPUs) and manufacture them itself (foundry). A laudable ambition, but a flawed model in the face of increasingly specialized competition, which prevented it from adapting to market upheavals. As a result, it missed decisive turning points, notably that of artificial intelligence. The central questions remain: can it still get back in the game? And if so, in what position?

Where is Intel going?

Intel's two divisions have different objectives and different opponents. On the CPU side (the processors that are the brains of our electronic devices), pressure is coming from all sides: ARM architectures are gaining ground at Qualcomm with mobile devices, Apple with its latest M4 chips, and even Nvidia with its Grace chips dedicated to inference. And on x86 architectures (the thought process of a processor, the standard for PCs since the 1980s), Intel has lost its grip. AMD, with its EPYC processors, has taken the throne and has no intention of letting go anytime soon.

Intel Foundry has made its ambitions clear: to become the world's second-largest foundry by 2030, just behind TSMC. This goal seems ambitious, to say the least. In 2024, TSMC generated $96.5bn in revenue, while Intel Foundry plateaued at $4.67bn, behind SMIC ($7.9bn), UMC ($7.7bn), GlobalFoundries ($6.75bn) and probably even Samsung. To bridge this gap, Intel is betting everything on its future etching technology (the manufacture of circuits inside a chip) to finally regain its competitiveness.

Technical point

Ambition is good. The means to achieve it is better. In this area, Intel is not so far off the mark... at least on paper. The group has fallen behind in etching processes and needs TSMC's production for its most advanced processors. This apparent weakness nevertheless allows it to focus its efforts on its own technology: the 18A process.

Comparable to TSMC's future 2 nm etching process, 18A has a technological advantage: PowerVia, a "rear" power supply system that promises better distribution of the chip's computing load and increased energy efficiency. This is a technology that TSMC does not plan to introduce before 2027 with its 16A node. In short, Intel could regain a head start. Microsoft and Mediatek have already expressed interest, provided that the product delivers on its promises.

For the chip design and marketing division, it's a different story. Although this unit has decision-making and financial autonomy, it remains heavily dependent on the technology schedule of the foundry division. The success of the upcoming Panther Lake (for PCs) and Clearwater Forest (for servers) ranges depends largely on Intel Foundry's ability to meet its market plan.

The risks

  • Well-established competition: Intel has lost valuable customers in both the professional and consumer markets. AMD reigns supreme in multitasking, while ARM continues to gain market share. Is there still room for Intel?
  • A place to be in AI: the AI giants (Nvidia, Microsoft, Google, Amazon, etc.) already have their supply chains in place. Intel, with no major customers, is struggling to find a place for itself.
  • Risky technological bets: between the accumulated delays and uncertainties surrounding the performance of the 18A process, the pressure is immense. The slightest deviation in production or schedule could prove very costly.
  • An unexpected change of direction: still marginal, the GPU division could become more important if Intel decides to accelerate in this direction. The mysterious Jaguar Shores chip could be the starting point. But that would not be the solution.

The expectations

  • Stabilize the Data Center segment: Data center revenues are erratic. A sustained growth trajectory over several quarters would be reassuring. This will notably require the ramp-up of the Gaudi range, dedicated to AI.
  • Improve margins: with an EBITDA margin of 20%, Intel is far from the industry's profitability standards (TSMC, Nvidia, and ARM are around 50%). A turnaround is crucial to regain market confidence.
  • Sign a major customer for Intel Foundry: Microsoft could be the first major customer for the 18A process. A concrete signing would be a real positive signal.
  • Consider a sale or spin-off: if the 18A technology fails, the Foundry division could be sold or spun off. This is a drastic option, but not unimaginable.
  • Meet deadlines: in an industry where trust is based on execution, sticking to the planned roadmap would demonstrate seriousness and credibility.

Intel seems to be on the road to recovery, but it cannot afford to make any mistakes. The company is moving forward in a precarious balance, and the slightest misstep could cost it dearly, especially in a sector as competitive as artificial intelligence. In this context, we prefer to remain cautious on the stock until Intel meets some of our criteria. It remains to be seen whether the plan will be executed. At the helm since March, Lip-Bu Tan embodies this hope: his time at Cadence Design left him with a reputation as a leader capable of turning things around.