Indium phosphide (InP) is emerging as the most fragile link in the AI infrastructure expansion chain. As data center demand for optical interconnects grows by orders of magnitude, this irreplaceable compound semiconductor material faces a supply gap that could be more severe than the current memory market turmoil.
Lumentum CEO Michael Hurlston issued a clear warning at the RAISE Summit in Paris this month: the shortage of InP will be more severe than the crisis currently experienced in the memory market. He noted that Lumentum currently operates five InP wafer fabs, but shipments still fall more than 30% short of customer demand, and this gap has widened from the previous quarter. "I don't think the two of us combined can meet the demand NVIDIA and other customers are placing on us," Hurlston said. "The InP shortage will be more severe than what's happening in the memory market."
Market research firm SemiAnalysis recently issued a similar warning, highlighting that supply of continuous-wave distributed feedback (CW DFB) InP lasers for near-package optical interconnects is particularly tight. The industry's wafer size remains predominantly at 2 to 4 inches, far from the 12-inch processes of silicon-based CMOS, presenting structural constraints on capacity expansion.
These warnings have sparked significant market attention. In March, NVIDIA invested $2 billion each in Lumentum and Coherent to secure capacity from these two suppliers, which together control most of the global high-speed data communication laser production. Despite this, related stocks have experienced substantial sell-offs over the past month: Lumentum (LITE) fell 22.3%, Coherent (COHR) dropped 37.6%, substrate supplier AXT (AXTI) declined 39.9%, and transceiver maker Applied Optoelectronics (AAOI) fell 35.7%, with stock price movements clearly diverging from fundamentals.
Silicon photonics cannot bypass InP
InP's irreplaceability stems from its material physics. Silicon's energy band structure is indirect, meaning it can only guide, split, and modulate light, but cannot emit it. InP's direct bandgap of approximately 1.34 electron volts allows it to efficiently convert electrical energy into photons. Therefore, whether it's NVIDIA, Broadcom, Marvell, or Cisco's silicon photonic platforms, all must incorporate InP lasers as light sources within the package.
The transition from pluggable transceivers to co-packaged optics (CPO) architecture changes the laser's location, not its removal from the bill of materials. In fact, the CPO shift pushes laser types toward more difficult-to-manufacture high-power continuous-wave sources and external laser modules to drive multiple optical channels. Coherent's partnership with NVIDIA specifically covers these high-power CW lasers, external laser source modules, and fiber array units.
NVIDIA claims its photonic switches reduce the number of lasers per port by four times, improve power efficiency by 3.5 times, and enhance network resilience by ten times compared to equivalent pluggable solutions. However, these savings are calculated on a "per port" basis, and it is the explosive growth in total port count that is driving the demand curve steeply upward. The high-end configuration of Spectrum-X Photonics supports 512 800Gb/s ports, with a total switching capacity of 400Tb/s.
A quantum leap in demand scale
Hurlston used a set of comparative figures to vividly illustrate the magnitude of this demand revolution: telecom customers order lasers in the "hundreds," while NVIDIA and hyperscale cloud providers demand in the "hundreds of millions." "Scaling from thousands of wafers to millions, and on non-silicon materials, is no small feat," he said.
Lumentum reported fiscal third-quarter revenue of $808.4 million, up 90% year-over-year, with component business revenue of $533 million and pump laser shipments growing 80% year-over-year. Yet even so, the supply-demand gap is widening. Hurlston stated on the earnings call that the imbalance has expanded from the previously disclosed 25% to 30% to "more than 30%," with pump laser supply tightness "even exceeding expectations," and core components "essentially sold out for the foreseeable future." The company's fourth-quarter guidance calls for revenue between $960 million and $1.01 billion, with operating margins expected at 35% to 36%.
Coherent also reported a record quarterly revenue of $1.8 billion, up 21% year-over-year, with data center and communications revenue share rising from approximately 41% a year ago to 75%, and a backlog extending into 2028. CEO Jim Anderson stated that the company's 6-inch InP line has achieved device yields surpassing its traditional 3-inch line, with internal capacity set to double by the end of the June quarter, one quarter ahead of schedule, and then more than double again by the end of 2027.
Duration of the shortage: Key differences from the memory crisis
InP supply tightness is not just a capacity issue; it is also constrained by structural raw material limitations and geopolitical risks. Indium, as a byproduct of zinc smelting, cannot be produced independently of zinc's economic logic, regardless of how strong the demand for lasers becomes.
Despite persistent warning signals, there is disagreement on the duration of the shortage. LightCounting's April 2026 market forecast shows a current transceiver demand exceeding supply by about 30%, consistent with Lumentum's data. However, the firm also expects the shortage to ease by the end of 2026, lowering its growth forecast for Ethernet transceivers this year from 82% in 2025 to 65%. Coherent's capacity doubling one quarter ahead of schedule also points in the same direction.
The key difference from the memory crisis is that InP's resolution path involves a wafer size migration already running on production lines with yields surpassing older node processes, rather than building a new fab from scratch, which would take three years. This supports a relatively optimistic timeline, but Hurlston's warning also has a solid basis in reality—his company's valuation depends significantly on how long this shortage persists.
Comments