Экс-глава Intel Пэт Гелсингер хочет перезапустить закон Мура с помощью световых чипов

Former Intel CEO Pat Gelsinger is betting on photonics — chips in which data is carried by tiny beams of light: this is how he hopes to restart Moore's Law and open the way to more powerful artificial intelligence, Wired reports. Gelsinger led Intel from February 2021 to December 2024 and gave the company more than 30 years of his career overall.

What is Gelsinger proposing?

Gelsinger proposes replacing electrons with light where classical silicon electronics runs into physical limits. He is talking about silicon photonics: signals inside a chip and between chips travel not along copper traces but through optical waveguides — those very "tiny beams of light." Optics carries more data with lower energy losses, and it is precisely this, by Gelsinger's logic, that can give the industry back the pace of growth that Moore's Law set for decades.

The idea is not new — fiber optics has long connected servers and continents — but now the point is to bring light inside the computing system itself: into the chip package and onto the die. It is this level that the industry considers the next frontier for AI hardware.

• Pat Gelsinger was Intel's CEO from February 2021 to December 2024 and previously the company's first chief technology officer
• Moore's Law was formulated by Intel co-founder Gordon Moore in 1965: the number of transistors on a chip doubles roughly every two years
• Gelsinger's new bet is photonics: transferring data inside chips and between them using light
• After leaving Intel, Gelsinger became a general partner in 2025 at the venture firm Playground Global, which invests in "deep" technologies

Why has Moore's Law stalled?

Moore's Law is an empirical rule that Gordon Moore derived in 1965: transistor density doubles roughly every two years, and computing gets cheaper. In recent years the rule has been working worse and worse: each new process node costs more than the previous one, the physical dimensions of transistors are approaching atomic scale, and heat output is growing faster than performance. Moore himself acknowledged during his lifetime that no exponential lasts forever.

Gelsinger knows these limits not from textbooks: under him, Intel tried to regain technological leadership in chip manufacturing, chasing Taiwan's TSMC and Korea's Samsung. Now, as Wired writes, the company's former chief sees the next leap not in further transistor densification but in the shift to optics.

"Pat

Gelsinger wants to pave the way to ever more powerful artificial intelligence with tiny beams of light," the Wired article says.

What does artificial intelligence have to do with it?

Training modern AI models takes place in data centers on tens of thousands of accelerators, and the bottleneck is increasingly not an individual chip but the links between chips. The larger the cluster, the greater the share of energy and time spent moving data rather than on the computations themselves. Photonic interconnects promise to remove this overhead: an optical channel carries more bits per watt than a copper one and works over longer distances without signal degradation.

Interest in photonics is shared by the industry's current leaders as well: the largest chipmakers and dozens of startups are running their own optical projects. For Gelsinger, who now looks at the industry through the eyes of a venture investor at Playground Global, this is at once a technological mission and a market opportunity: whoever is first to bring photonic chips to mass production will hold the key infrastructure of the AI era.

What does this mean?

If the bet on light pays off, Moore's Law will get a second wind, and AI's computing ceiling will be pushed back without exponential growth in energy costs. Gelsinger's voice carries weight here: when an engineer who spent three decades building classical processors publicly bets on photonics, the industry listens.