ABF substrate TAM is driven by AI GPU die area growth, which is compounding faster than substrate capacity can expand. Substrate makers can add 10–15% capacity per year; NVIDIA GPU die area is doubling every 1.5–2 years. The base case reflects a sustained supply shortage through 2027 followed by gradual capacity normalization as Taiwanese substrate makers and Ibiden complete expansion investments. Bear case assumes glass core substrates achieve volume production faster than expected (Intel targeting 2028), begin qualifying at AMD and Broadcom, and reduce ABF film demand as a proportion of total substrate content. Bull case assumes glass substrate qualification delays push full adoption past 2030, AI GPU die area continues compounding, and ABF film remains the sole viable substrate material through the forecast period — sustaining Ajinomoto's pricing power.
“Taiwan's ABF substrate capacity is fully booked for 2026. Spot prices have jumped 30–40% since Q1 2026 as AI GPU demand overwhelms supply. The supply-demand gap is now expected to extend through 2027. Unimicron, Kinsus, and Nan Ya PCB — Taiwan's three major ABF substrate makers — all guided sold-out capacity for the year.”
“GPU substrate area is compounding demand: NVIDIA Hopper required 3,190mm² of ABF substrate per chip, Blackwell requires 5,913mm², and Rubin is estimated at 8,051mm². Each generation requires ~2× the substrate area of the previous, while substrate suppliers can only expand capacity 10–15% per year. The math does not close without either new capacity or chip architecture changes.”
“Ajinomoto — the Japanese food company that invented ABF film and holds near-monopoly supply of the base material — announced a 30% price increase for its ABF film effective H2 2026. Ibiden, the largest ABF substrate maker, disclosed it has secured enough ABF film for current guidance but "not yet secured materials to accommodate potential upside." Upstream material constraint is binding before substrate capacity constraint.”
“Intel and Samsung are actively developing glass core substrates as an alternative to ABF-based organic substrates. Glass substrates offer better flatness, lower signal loss, and higher interconnect density. Intel is targeting 2028 for first glass substrate production at scale. If successful, glass substrates begin to reduce Ajinomoto's effective monopoly on the film layer — but qualification timelines at major chipmakers make pre-2029 adoption unlikely.”
“Beyond GPUs, ABF substrates are required for high-end CPUs (AMD EPYC, Intel Xeon), networking ASICs (Broadcom Tomahawk, Marvell Teralynx), and AI accelerators from every hyperscaler. The AI chip rivalry across NVIDIA, AMD, Broadcom, and custom silicon is pulling substrate demand from multiple directions simultaneously — unlike prior cycles driven by a single product category.”
NVIDIA's substrate area requirement has grown 2.5× in two generations and is on track for 8,000mm² with Rubin. No substrate maker can double capacity in 18 months — capacity investments take 3–5 years from decision to volume production. This structural mismatch between AI GPU roadmap cadence and substrate capacity expansion creates a permanent supply gap as long as NVIDIA's die area trajectory continues.
Ajinomoto invented ABF film in the 1990s and holds an effective monopoly on the base material from which all ABF substrates are made. There is no commercial-scale alternative film today. The 30% price increase in 2026 is the first significant exercise of that monopoly pricing power — and it flows directly through to substrate makers and ultimately to chip customers. Ajinomoto's constraint is upstream of the substrate constraint: even if Ibiden or Unimicron wanted to expand faster, they cannot without more ABF film.
Intel and Samsung are developing glass core substrates that could replace ABF film entirely. Glass offers superior flatness (critical for sub-2nm chip bonding), lower electrical loss at high frequencies, and potentially higher interconnect density than organic ABF substrates. Intel is targeting 2028 for production-ready glass substrates. If the technology qualifies at AMD, Broadcom, or NVIDIA before 2030, it disrupts Ajinomoto's monopoly and caps ABF substrate pricing power.