The market will read Altera's return to Wall Street as an AI story. It is arriving as one of the largest semiconductor listings since Arm's 2023 debut, at a moment when U.S. IPOs are running at a record pace. Before the AI narrative, however, sits an accounting fact worth anchoring the whole read on: the exact same asset was written down to roughly half its prior value a little over a year ago. That is not an AI company's usual pre-IPO posture.
Altera, the programmable-chip maker IntelINTC-- acquired for $16.7 billion in 2015, is preparing to file confidentially for an initial public offering that could raise more than $2 billion, with a listing possible as early as this year. Silver Lake, which controls a 51% stake, has tapped Barclays, Citi, JPMorgan, and Morgan Stanley as underwriters. Intel retains the remaining 49%.
The value arc, not the headline, frames the deal
The sequence is the frame. Altera was bought by Intel in 2015 for $16.7 billion, largely vanished into a CPU-centric sales machine, and resold in September 2025 — after a planned IPO was canceled — when Silver Lake took 51% for $4.46 billion, valuing the whole company at $8.75 billion. That is the undeniable baseline this offering builds from.
Quick Backtesting Tool
| Step | Year | Event | Value |
|---|---|---|---|
| Intel buys Altera | 2015 | Acquisition | $16.7 billion |
| Altera revenue | 2023 | Sold separately by Intel | $2.9 billion |
| Altera revenue | 2024 | Halved | ~$1.5 billion |
| Silver Lake buys 51% | Sept 2025 | $4.46 billion paid, Intel keeps 49% | $8.75 billion |
| IPO | 2026 | Confidential filing, $2B+ raise | Pitched "meaningfully above" $8.75B |
(Compiled from public sources; Altera has disclosed no detailed figures since turning private.)
Why did the value collapse? The FPGA market did not simply soften — it split into two segments with very different economics. During 2023–2024, data-center and AI budgets shifted decisively toward GPU accelerators and custom ASICs, draining spending from programmable logic, while Altera simultaneously lost share to AMD-owned Xilinx. Altera's revenue fell about 50%, from $2.9 billion in 2023 to roughly $1.5 billion in 2024. AMD's embedded segment, which is mostly the former Xilinx, fell a third in 2024, to $3.6 billion. Both leaders shrank together, which is the signature of a demand contraction in a category being squeezed at the margin, not a company being picked apart.
The rebound is real, but it is off the trough
Under Silver Lake and new CEO Raghib Hussain, Altera reports it grew revenue "roughly 20%" last year and expects mid-20s percent growth this year, with operating income more than doubling. Two cautions belong next to those claims. First, that growth is measured off a base that was just halved; a snap-back from a 50% revenue collapse carries very different implications than an expansion from a steady state. Second, the company is private and has published no details — so the unit-versus-price split behind the growth is unverified.
The supporting evidence matters, because it lets us trust that the recovery is a real category event rather than Altera alone. The one public pure-play FPGA stock — Lattice Semiconductor, the low-power niche player — grew revenue 42% year over year in Q1 2026 and guided to roughly 50% growth in Q2 2026; its shares are up about 55% year to date and roughly 77% over the last twelve months. AMD's embedded segment also returned to growth. This is genuinely a sector snap-back, which is the strongest check we have on Altera's claims.
The structure of that recovery is the two-market split this listing will be priced on, and it runs the other way from the opening AI narrative. FPGAs lost the data-center and large-scale inference fight to GPUs and ASICs — that is where the 2024 damage came from. Their defensible ground is edge, robotics, aerospace and defense, networking, and industrial automation, where workloads are deterministic, real time, and reconfigurable. The current upswing is a rebound in that traditional base, not a breakout in data-center AI.
The projection to separate from the evidence
What requires the most scrutiny is the forward claim. Hussain describes FPGAs as the "nervous system" to the GPU's "brain" and projects FPGA content of $100 to several hundred dollars per robot, in a market he sizes at "$100 billion to several hundred billion" over a decade. That is aspiration, not evidence. The category contracted sharply immediately before the projection was made, and robotics machine volumes are not yet visible in Altera's shipment figures. No disclosed or reproducible numbers currently tie the robotics thesis to revenue.
Two structural facts further temper the valuation case. Altera's leading-edge products are moving to TSMC's 2-nanometer and 3-nanometer processes, so the fabrication constraint it lives under has migrated to a Taiwan-dependent foundry — a concentration risk for a vendor that sells into U.S. aerospace and defense programs. And the base economics remain thin: Altera posted a $615 million GAAP operating loss in 2024, with non-GAAP operating income of just $35 million on $1.54 billion of revenue. A customer designing Altera into a 10-year defense or robotics platform is now underwriting a newly independent, sub-scale supplier whose revenue halved a year earlier.
Where the reader's stake sits
For Intel holders, this is the payout leg of the capital-reallocation story that has lifted the stock from about $20 in early 2025 to roughly $100 today under CEO Lip-Bu Tan. Intel is divesting non-core assets and raised about $20 billion in August to fund manufacturing and AI ambitions; an Altera offering "meaningfully above" the $8.75 billion mark would be at least partial vindication for a stake it originally bought for nearly twice that. The IPO monetizes a holding Intel has already moved off its books at a loss.
For anyone eyeing the offering itself, the number to watch in the S-1 is not the topline growth headline — it is the split behind the 20–25% growth: how much is volume in the recovered traditional FPGA base, and how much is forward-priced AI and robotics content that has not yet shipped. The key issue is not whether Altera is an "AI company" at pricing. It is whether the two segments it now spans are being valued together at the growth rate of the second, the one still living in projections. When the listing settles, that is where the risk premium will either hold or unwind back toward the economics of a sub-scale, marginally profitable FPGA vendor with its fabrication tied to Taiwan.
