Altera's $2B+ IPO Forces a Hard Question: Is FPGA an AI Compute Asset or a Foundry Side-Show?

Altera was already a known name in programmable logic when Intel acquired it in 2015 for $16.7 billion. By April 2025, with Intel bleeding cash into its foundry build-out and FPGA market share having slid against AMD's Xilinx, Intel struck a deal to sell 51% of the unit to Silver Lake at an $8.75 billion valuation — roughly half what it paid a decade earlier. Abu Dhabi-backed MGX co-invested, and the transaction closed September 12, 2025, with Intel keeping49% and consolidating Altera's results under the equity method from that date forward.

Seventeen months later, Reuters reported on September 9, 2026, that Altera is preparing to confidentially file for an IPO that could raise over $2 billion as early as this year. Barclays, Citi, JPMorgan and Morgan Stanley have been tapped as underwriters, with Morgan Stanley reported as the frontrunner for the lead-left role. The filing will test a market question that has not been answered publicly since the carve-out: what is the right multiple for a stand-alone, mid-single-digit-percent-growth FPGA vendor whose book of business skews toward industrial, aerospace/defense and 5G networking?

Altera's last fully consolidated financial year — FY2024, when it was still reported as Intel's Programmable Solutions Group — generated $1.54 billion in revenue. GAAP gross margin was $361 million, or about 23%, weighed down by the manufacturing cost structure of Intel-owned fabs. After $976 million of GAAP operating expenses, the segment posted a $615 million GAAP operating loss. On a non-GAAP basis, which strips out stock-based compensation, amortization and certain impairments, gross margin was $769 million and operating income was a thin $35 million — a 2.3% operating margin.

The trajectory accelerated in the first half of 2025: $816 million of revenue at a 55% segment gross margin, on $356 million of operating expenses. That puts H1 2025 gross margin sharply above the FY2024 figure, partly because Intel's PSG report in H1 2025 reflects margin after intersegment allocations and partly because Altera was already shifting toward more profitable Agilex5/Agilex 3 mix. Operating leverage remains the central question: even at55% gross margin, the segment needs substantially higher revenue to absorb the R&D base a stand-alone FPGA business requires.

Altera's market position has slipped materially since the Intel acquisition. Gartner data cited across the industry put Xilinx at roughly 51% of FPGA revenue and Altera at 29% in 2021, with Lattice Semiconductor at about 7% and Microchip Technology at 6%. The Altera share number dropped further through the Intel ownership era as engineering focus fragmented between data-center GPUs, foundry process nodes and FPGA development. A stand-alone Altera must claw that share back — the IPO valuation will reflect how much of that recovery the market believes is realistic.

Market-cap context for AMD and Microchip was inferred from price-per-share and outstanding shares in their respective earnings disclosures; check live quotes before trading.

Lattice Semiconductor is the cleanest pure-play comparable, and the multiple is rich: $15.8 billion market capitalization on $651M of trailing revenue is a 24x EV/Sales print — the kind of premium Altera could plausibly chase if the IPO narrative emphasizes AI-edge inference and physical-AI robotics rather than telecom and defense. Lattice Semiconductor itself is up roughly 150% over the prior year on that thesis, with February 2026 quarterly revenue of $146 million and Q1 2026 guidance of $158–172 million. The deeper FPGA-and-custom-silicon peer set — AMD and Marvell Technology — bundles FPGA into AI accelerator franchises that the market is valuing at10–15x forward sales, providing a natural ceiling.

Intel's urgency to monetize Altera is not a stand-alone corporate finance decision — it is a direct consequence of the foundry rebuild. The Intel Foundry segment posted a $2.089 billion operating loss in Q2 FY2026 on $5.77 billion of revenue, a 36% operating loss, after a $2.4 billion loss in Q1. Earlier quarterly losses ran deeper — $3.2 billion in one quarter of2025 — as18A and 14A ramp costs outpaced external customer revenue. Q2 FY2026 Intel Products revenue of $15.1 billion was up 28% year-over-year on13% growth in CCPG (Client Computing & Physical AI) and 59% growth in DCAI (Data Center & AI), but the foundry drag still turned the consolidated TTM result into a $11.3 billion net loss.

To fund the 18A ramp and 14A preparation, Intel has pulled together an unusually large capital stack: a $20 billion common stock offering priced on August 11, 2026 at $95 a share (upsized from $15 billion and closed August 12); $5 billion of equity from Nvidia at $23.28 a share earlier in2025 (giving Nvidia an estimated 4% stake); $2 billion from SoftBank; and a U.S. government equity position of approximately 9.9%. Every dollar of equity raised reduces the pressure on Altera as a captive revenue source inside Intel, and the IPO would convert Intel's retained 49% into a tradable mark-to-market position rather than a consolidated operating unit.

Upstream: Altera's silicon roadmap depends on Intel Foundry as its primary node supplier for Agilex 5 and the upcoming Agilex 9 family, which is both a strategic advantage and a concentration risk. Microchip Technology's FPGA business (PolarFire, IGLOO2) and Lattice Semiconductor (Avant, Certus-NX) both rely on TSMC and Samsung foundries, which means Altera's pitch to customers on \"secure onshore supply\" — repeated in the September 15, 2025 Silver Lake closing announcement — becomes a real differentiator if Intel18A hits its yield targets.

Downstream: Altera's customer base skews toward industrial automation, aerospace and defense, audio/video broadcast, robotics, telecommunications (5G base-station DSP and fronthaul), data-center networking and edge AI — the same end-markets Silver Lake cited in its closing press release. The AI-edge pitch leans on the FPGA AI Suite 2026.1, which adds spatial compilation for physical-AI inference workloads on Agilex 5 and Agilex 3. Comparable peers AMD (Versal AI Edge Gen 2) and Lattice Semiconductor (sensAI 8.0) are pursuing the same workloads; the IPO will reveal whether Altera's installed-base breadth in industrial and defense converts into AI inference design wins.

Short-term (days to quarters): the next confirmed inflection points are the confidential S-1 filing with the SEC (expected in coming weeks per Reuters), the public S-1 with disclosed price range, and the IPO pricing itself. Each milestone will force the market to assign a multiple to Altera's $1.54B run-rate revenue and its sub-3% non-GAAP operating margin. The cleanest read on investor appetite will come from comparing the implied IPO valuation to the $8.75B Silver Lake deal mark — a flat-to-up outcome would confirm the FPGA re-rating thesis; a sharp discount would signal the market views the stand-alone business as less valuable than Intel's captive structure. Quarterly Altera disclosures will not be public until after the IPO, so near-term tracking will rely on supply-chain signals (wafer orders from Intel Foundry, design-win announcements in edge-AI robotics) and on Intel's continued equity-method reporting of its 49% stake.

Long-term (1–3 years): the strategic question is whether Altera can re-take FPGA share lost to Xilinx during the Intel years while expanding into AI-edge inference and physical-AI robotics. Three measurable milestones will decide: (1) Agilex 9 tape-out and volume production on Intel 18A — proof that the captive-foundry model survives the deconsolidation; (2) FPGA AI Suite design-win traction versus AMD's Versal AI Edge Gen 2 and Lattice Semiconductor's Avant/Nexus platforms; (3) the path to GAAP operating profitability, which given FY2024's $615M loss and $361M GAAP gross profit requires either a 2–3x revenue lift or a sustained50%+ gross margin. The biggest risk to the thesis is Intel itself: if 18A yields disappoint or 14A delays force further equity raises, Altera's \"onshore secure supply\" pitch loses its primary differentiator and the IPO multiple compresses back toward legacy semiconductor norms.

Altera's IPO is the cleanest live test of whether programmable logic gets re-priced as an AI compute asset. The setup is unusually clean: a stand-alone business with disclosed segment financials, a defined AI-edge product roadmap, a captive U.S. foundry supplier, and a recent sponsor mark at $8.75B. The risk is equally clear: FY2024 GAAP operating loss of $615M and a non-GAAP operating margin under 3% leave no cushion for execution missteps, and AMD's Xilinx franchise — bundled inside a chipmaker that booked $34.6B of FY2025 revenue and trades at the multiple of a hyperscaler-adjacent AI accelerator designer — sets a hard ceiling.