THE LATEST NEWS
Intel Sells Majority Stake in Altera to Silver Lake

Ailing chip giant Intel has sold a majority stake in FPGA maker Altera to private equity group Silver Lake. The deal values Altera at $8.75 billion, losing about half of its value since it was acquired by Intel for $16.7 billion in 2015.

Intel announced it would spin out Altera in February as previous leadership sought to cut the company’s costs. At the time, Altera said it was seeking an investor to take a minority stake, and that it was considering an IPO for 2026. The news that a majority stake in Altera will be sold off fits with new Intel CEO Lip-Bu Tan’s plan to spin off non-core businesses as the chipmaker struggles to finance its foundry play while making a mark in the burgeoning data center AI space.

“Today’s announcement reflects our commitment to sharpening our focus, lowering our expense structure and strengthening our balance sheet,” said Intel CEO Lip-Bu Tan in prepared remarks. “Altera continues to make progress repositioning its product portfolio to participate in the fastest growing and most profitable segments of the FPGA market.”

In 2024, Altera generated revenues of $1.54 billion against an operating loss of $615 million; it had around $2 billion in annual revenue when it was acquired.

Intel will retain 49% of Altera with the other 51% being owned by Silver Lake going forward.

Generate value

Intel acquired Altera 10 years ago against a background of consolidation in the then-mature (pre-AI) semiconductor industry, marked by a string of M&A deals including Avago-Broadcom and NXP-Freescale. Altera had around $2 billion in annual revenues at the time of its acquisition.

Analysts EE Times spoke to at the time suggested Altera was over-valued, and that Intel would struggle to get value out of the FPGA maker, given a number of factors. These included the limited opportunity for manufacturing Altera products in Intel’s fabs, the fact that Altera was already run as a pretty tight ship with minimal space for cuts to improve operational efficiency, and Intel’s already terrible track record of integrating acquired companies.

At the time of the acquisition, then-Intel CEO Brian Krzanich said the company would ship co-packaged CPU/FPGA products for the data center and embedded markets, followed by products that merged the two types of processor on the same silicon.

An Intel Xeon Skylake CPU with co-packaged Altera Arria 10 FPGA was produced in 2018 but never productized; it was intended to form Intel’s data center roadmap alongside Omni-Path interconnect (cancelled in 2019), Nervana AI accelerators (cancelled in 2020) and Optane 3D X-point memory (cancelled in 2022).

“The synergies between FPGAs and data center microprocessors never appeared,” Joe Byrne, principal at Xampata Insights, wrote on LinkedIn. “Nor did the synergies between Altera’s products and Intel’s manufacturing might, which began its downward slide after the two combined. Meanwhile, investment in the Altera product line languished. More recently, the catch-all industrial semiconductor market, telecom market and China market—major FPGA segments—have faced cyclical and geopolitical headwinds. Thus, it’s a tough time for Altera to spin out, but Intel needs cash and operational focus.”

“Factoring in the above along with Intel’s acquisition track record, 50% value loss isn’t bad,” Byrne added.

Product roadmap

Altera’s core FPGA products fall into three tiers, whose brands have been unified under the Agilex brand in recent years. They were the high-end Stratix range, mid-range Arria products and low-end Cyclone and Max FPGAs targeting cost-sensitive applications. Altera was already using Intel’s fabs for its high-end Stratix 10 FPGAs (on Intel’s 14-nm process) and Intel’s advanced packaging technologies for its multi-die products.

However, EE Times reported in November 2024 that post acquisition, Altera teams struggled to use Intel-proprietary EDA tools that were optimized for making CPUs. CPUs, characterized by their dense logic and minimal interconnect, are the opposite of FPGAs’ balance of an extensive network-on-chip (NoC) connecting sparser logic and memory blocks.

The latest parts in Altera’s Agilex 7 range began production on the Intel 7 process node about two years behind competitor Xilinx’ 7-nm Versal, following delays to Intel’s process technologies, possibly combined with lack of priority access to Intel’s process technologies for Altera. Meanwhile, Altera and Xilinx face increased competition at the low-end of the FPGA market from Lattice and Microchip, an area where neither of the leaders has introduced new products in a decade.  

Altera CEO Sandra Rivera told EE Times at Embedded World last month that being a part of Intel had offered the best of both worlds to the company.

“Intel is a wonderful company, lots of innovation, lots of brand, reach and scale, and the ability for us to leverage that has been a positive for certain parts of our portfolio, certainly the high end of the portfolio, the work we’ve done [in] communication infrastructure and data center and cloud,” she said. “But the ability to spin out and leverage all that goodness from Intel, to recommit and refocus in the mid-range and lower end of the overall stack […] is the best of both worlds—we get to operate a lot more independently, with more speed and agility, we get to partner with our ecosystem and our channel […] but with all of the backing and support from Intel.”

Like competitor Xilinx, Altera is trying to reposition itself for the age of AI, the single factor driving semiconductor growth across the board. Rivera said at Embedded World that the mid-range Agilex 5, launched last year to ramp this year, is well-positioned for AI as it includes DSPs with new AI acceleration capabilities as part of its compute fabric.

Altera’s latest product launch, the Agilex 7-M, is a high-end FPGA with hardened memory NoC to improve HBM bandwidth for data center AI inference applications (this is already used by AI startup Positron with impressive results).

New CEO

The deal with Silver Lake will re-establish Altera as an independent company—it will become the largest pure-play FPGA company since market leader Xilinx has since been acquired by Intel rival AMD.

As part of the deal, Rivera will be replaced as Altera CEO by Raghib Hussain, effective May 5. Hussain is currently president of products and technologies at Marvell, which he joined in 2018 when the company he co-founded, Cavium, was acquired. Hussain had served as data center networking chip company Cavium’s CTO and latterly COO through the company’s IPO and later acquisition.

Silver Lake has a track record of investing in semiconductor companies. The private equity firm together with KKR acquired the semiconductor division of Agilent in 2005 and renamed it Avago, later investing further in the company to acquire LSI Logic and Broadcom. In 2006, Silver Lake was part of a consortium that acquired a controlling share of Philips’ semiconductor division, which became NXP Semiconductors.

A statement from Kenneth Hao, chairman and managing partner of Silver Lake, indicated the company intends to keep at least some of Altera’s business with Intel Foundry going forward.

“This investment represents a once-in-a-generation opportunity to invest in a scale leader in advanced semiconductors. Together with Raghib, we will be focused on strengthening Altera’s technology leadership position and investing in emerging AI-driven markets such as edge computing and robotics,” Hao said. “We look forward to working closely with Intel as a strategic partner who will continue to provide U.S.-based foundry services and complementary engagement with customers.”

The transaction is expected to close in the second half of 2025.

From EETimes

Back
Intel Sells Majority Stake in Altera to Silver Lake
Ailing chip giant Intel has sold a majority stake in FPGA maker Altera to private equity group Silver Lake. The deal values Alter...
More info
Arm Powers Software-Defined Vehicle Revolution
The automotive industry, long defined by combustion engines and mechanical prowess, is undergoing a seismic shift. Arm is playing a centra...
More info
How Did Nvidia Improve Hopper Inference Performance 30x?
SAN JOSE, Calif.– Nvidia has dramatically boosted the inference performance of its GPUs with new data center inference orchestr...
More info