Three years after becoming Intel CEO, Pat Gelsinger has detailed his vision at a special event in San Jose, California, to mark the official launch of Intel Foundry, as part of plans to build AI chips for itself and other companies. The purpose of the move is twofold – to attract the major chip designers, such as Nvidia, AMD, Qualcomm Apple, Amazon and Microsoft and become the world’s second largest chip manufacturer after Taiwan’s TSMC, and to prove that the company wants to once again become the number one chip manufacturer in the US.
There are quite a few obstacles on the way to realizing the company’s vision for the coming years, which also pose a challenge for Intel in Israel, where it is the country’s largest private employer.
Intel has been manufacturing chips on a limited scale for companies like ARM and even Nvidia for several years but now Gelsinger is separating manufacturing and development to prevent structural conflict of interests with manufacturing becoming an independent profit and loss unit
Intel Foundry will include IFS, the fastest growing production unit in Intel in 2023 with revenue of nearly $1 billion. The division will include all the group’s existing manufacturing fabs worldwide such as the chip plants in Kiryat Gat and its flagship plant in Oregon, which is responsible for proving the programming of new production technologies.
The new division will specialize in services for external customers and developing future production processes. Intel’s development centers will work like Nvidia and ARM as if there is not a fab serving all their whims, but a plant that manufactures for customers according to price, works on development plans in a thorough manner and preempts making future orders. Thus, Intel’s development engineers could find themselves producing at rivals such as TSMC and Samsung. Intel believes, even if they do not say so explicitly, that the company’s added value is no longer found in its core business – the development and production of chips for personal computers – a shrinking market with marginal profitability, which is decreasing over the years – but in its production system. Thus, Intel aspires to be one of the world’s largest AI chip manufacturers, rather than developing the most successful AI chip itself.
The move, which has been planned years in advance, is branded by Intel as one that will make it the main manufacturer for AI chips. This is a cultural and business earthquake within Intel, and the biggest organizational change the company has undergone since its founding in 1968. In the past, Intel dominated the global chip market: first, thanks to the PC revolution in the 1980s and then, thanks to the laptop revolution in the 1990s and early 2000s, partly thanks to chips developed in Haifa.
RELATED ARTICLES
Intel confirms further $15b Kiryat Gat fab investment
The past decade has been particularly tough for Intel, which has lost ground in several areas. It has failed to penetrate the smartphone market, while most AI chips currently produced for servers and data centers are developed and manufactured by rivals. Because of its closed and unique structure as a chip company, which produces only what it develops, without being open to external developers, its corporate culture has undergone a certain degeneration, and its ability to innovate and lead complex production processes has been weakened. Thus, almost all the main competitors have overtaken Intel in terms of market value and momentum in the capital market. Nvidia and AMD have overtaken it in market value and chip share in the AI server market.
Intel’s great dependence, which remains, on chips for PCs, makes it difficult for it to show a significant improvement in financial results and accumulate enough cash to finance its new strategy. Relying on government grants and partner-funded collaborations provide Intel with an oxygen pipeline in the meantime, with the aim of reaching the long-awaited pivot.
How does Intel plan to overtake rivals
In the special launch press conference, Intel set out to demonstrate that it is ready, for the first time, for customers in the manufacturing field and that it is very serious about its production plans. The conference was held in partnership with senior industry figures like OpenAI found and CEO Sam Altman, Microsoft CEO Satya Nadella and US Secretary of Commerce Gina Raimondo.
Unlike its competitors, such as TSMC and Samsung, which operate in secrecy and are ambiguous about their production plans, Intel has set out a very detailed roadmap, which may give it a certain advantage in production for possible key customers such as Nvidia, Apple and AMD.
While until recently, Intel was a generation or two behind TSMC in chip architecture, the company is announcing the development of manufacturing capability for 1.4 nanometer (14 angstrom) chips. This is the smallest size transistors – the electrical circuits that form the basis of the chips – that has been announced by a chip manufacturer so far, with an expected launch in 2026.
This is an ambitious plan that is raising eyebrows in the industry, with a leap of several generations at once for Intel. So far Intel has only mass-produced chips with technology up to 7 nanometer only and lags behind TSMC, which produces 5 nanometer chips – which have been embedded in Apple devices and Nvidia’s data centers for over a year.
TSMC and Samsung are already close to developing 2 nanometer chips, while Intel’s 5 nanometer technology (known as Intel 3) has just completed development and may enter production soon. Intel promises that in 2024, advanced chips with 2 nanometer (20 angstrom) and 1.8 nanometer (18 angstrom) technology will be launched for mass production. The company claims that its chips will be more efficient than competitors, and more adapted to processing AI and working in data centers, through an architecture that allows power to flow to transistors from many directions, allowing more efficient energy consumption and savings in electricity expenditure.
Intel also has advantages that its rivals don’t possess, which may enable it to prevail, even over TSMC, in the long term. Intel has invested in a network of chip assembly, packaging and testing plants in New Mexico, Costa Rica, Malaysia and Poland – an activity that is almost non-existent among its competitors and is considered critical for complex chip production from several processors. Intel hopes to use these assembly plants to integrate its own processors into chips that are ordered and designed by customers. The chips include a collection of processors in various fields – memory, encryption, core processing, graphic processing and AI, from different manufacturers, to order.
Intel also has a special connection with Dutch company ASML, the world’s only company manufacturing machinery for producing advanced chips and which provides it production devices for all advanced chips. This, in contrast to TSMC, which makes more limited use of them and relies more on internal production. Intel has founded a network of extensive connection with chip design companies like Cadence, Synopsis and ARM, formerly a bitter rival, to allow startups and growth companies to produce their chips on its production lines. Intel was also informed this week that it will receive the highest grant given so far to a US chip company based on the chip law – $10 billion, an amount equivalent to the construction of half a factory. For comparison, Global Foundry received only $1.5 billion.
Intel is making the change after a period of downsizing and restructuring in some divisions, although the company did not implement a massive wave of layoffs as it did in 2016, when it dismissed 12,000 employees. In 2022, the company implemented pay cuts that ended last year, when bonuses were also restored.
Nevertheless, Intel faces obstacles on its ability to realize its ambitious vision. The first is whether it can attract major rivals such as Nvidia, Qualcomm and AMD, as well as past customers, like Apple and Amazon. These companies have become accustomed to working closely with TSMC, and although some of them may produce at Intel, we will never know what production volume it transfers to it.
In a situation where demand for AI chips is so big, any factory that produces may win contracts. But competitors may fear that their commercial secrets will leak into Intel’s development departments. The company guarantees a sealed wall and zero transfer of information, but at the same time, they have avoided completely spinning off production activity as analysts recommended. Such a split, no doubt, would send Intel stock higher, and every day that goes by without spinning off the manufacturing plants leaves Intel stock lower.
Gelsinger’s decision to keep the production plants within the company has a certain business logic. Most of the company’s cash still comes from the PC development business, although Intel’s cash flow, as a whole, continues to break negative records.
There is also the question of whether the high demand for AI chips will continue growing. Today, there is a high shortage of AI chips, due to the monopoly of Nvidia in development and TSMC in production. The industry fears a flood of chip factories after a massive opening of factories in the coming years, as well as a lack of suitable workers, which has already led to delays in the construction of the factory in Magdeburg in Germany. Intel will have to hire tens of thousands of production workers and engineers in places like Germany, Ohio, New Mexico and Israel, and there are difficulties in finding them and transferring them to Intel.
The Israeli angle
As part of the reorganization, Intel’s existing plant and future plant in Kiryat Gat will join the new independent production division, and employees at the plant are now undergoing training to service external customers. However, it has not yet been decided in what way, for which customers and which technologies will be formed for them in Israel.
The Fab 28 plant in Kiryat Gat currently produces silicon wafers for chips with 10 nanometer technology. However, Israel is the only site in the world outside of the US where there is very close proximity between Intel and Nvidia facilities, and the company is considering how to use this proximity to attract Nvidia to them, with Israeli assistance. In addition, several Israeli companies have already signed on to manufacture chips for Israeli startups – including Valens Semiconductor (NYSE: VLN) and Ceva (Nasdaq: CEVA), using Israeli companies at TSMC’s expense is a refreshing breath of fresh air.
Israeli chip company NeuReality founder and CEO Moshe Tanach says, “This move in which Intel is reorganizing its production processes will take at least two to three years. We will see the first buds of the trials and tests by the end of the year and during 2025. This is a process that Intel has already tried in the past, but rigidity and conservatism on the part of the production units prevented it.”
Tanach added, “Although Gelsinger is placing a greater emphasis on this process, in a large company, which at the same time has to continue with the production of its original products, generating revenue and profits for the shareholders is a very challenging task – a challenge that arises from two vectors that sometimes collide with each other. At the same time, rivals like Nvidia and AMD have opened a gap on Intel’s capabilities in developing graphics processors as well as core processors but Intel is the only one that possesses an advanced production capability, which would be difficult if not impossible for other companies to develop.”
Full disclosure: The author was a guest of Intel in San Jose
Published by Globes, Israel business news – en.globes.co.il – on February 22, 2024.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2024.
