Semiconductor Materials, PFAS Alternatives, Green Methanol, Advanced Polymers, AI-Designed Catalysts: Five Material Directions Worth Tracking Over the Next Five Years

"New materials" has become one of those phrases that gets stretched to cover almost anything — graphene, AI drug design and green hydrogen all show up in the same paragraph of most industry outlooks. Laid out against what actually happened between 2024 and 2026, the picture is narrower and messier. Some directions are advancing roughly on schedule (Japan and Germany's real capital spending on silicon carbide, for instance). Others that got a lot of press quietly fell apart — a Chinese green methanol project once billed as the world's largest was cancelled this June with no public explanation. And some widely cited "breakthroughs" haven't held up under peer scrutiny, DeepMind's GNoME materials-discovery claims being the clearest case. What follows are five directions that still look solid once the hype is separated from what's actually been disclosed.

1. Semiconductor Materials: Wide-Bandgap Devices Are Scaling, Packaging Is Filling the Gap, 2D Materials Aren't There Yet

Silicon carbide is the most concrete part of this story. Japan's NEDO (New Energy and Industrial Technology Development Organization) disclosed in April that ROHM's 8-inch SiC MOSFET line at its Chikugo plant hit its Green Innovation Fund technical targets two years ahead of schedule; ROHM's own target is a 35x increase in SiC device capacity by fiscal 2030 versus fiscal 2021, backed by a new 200mm wafer plant in Miyazaki. Germany's Infineon opened phase one of its 200mm SiC fab in Kulim, Malaysia in August 2024; total planned investment across both phases runs €2–7 billion, underpinned by roughly €5 billion in customer design-win commitments and €1 billion in prepayments — Infineon's own stated goal is about 30% of the global SiC market within the decade. Market-size forecasts for wide-bandgap semiconductors disagree wildly between vendors (compound annual growth rates ranging from 10.8% to 26% depending on whose report you read), which is itself a sign the category is still being sized inconsistently — treat any specific number as directional.

On the lithography materials chain, ASML's first High-NA EUV systems are scheduled to enter series production starting this year, with the first module delivered to imec's 300mm cleanroom in Belgium in March. Germany's ZEISS SMT built the optics: the projection optics alone run to more than 40,000 parts and roughly 12 tons, the illumination system another 25,000-plus parts and 6 tons; about 2,000 of ZEISS SMT's 9,349 employees work on High-NA development, backed by some 10 million R&D hours and 2,000-plus related patents. Photoresist chemistry remains almost entirely Japanese — Shin-Etsu, JSR, Tokyo Ohka Kogyo, Sumitomo Chemical and Fujifilm together hold about 90% of the global market — and the industry is going through its first resist-chemistry overhaul in roughly thirty years, shifting from chemically amplified resist to metal oxide resist. JSR, through its Inpria acquisition, expects metal-oxide resist to start contributing revenue in 2025–2026; Tokyo Ohka Kogyo is targeting full commercialization for 2026–2027. Germany's Merck KGaA opened a roughly $600 million semiconductor-materials megasite in Kaohsiung, Taiwan last December, and separately announced an €85 million, three-year expansion of its EUV photoresist R&D center in Darmstadt, working with imec on non-CAR, metal-oxide-nanoparticle resists for sub-2nm nodes.

Advanced packaging is the other major thread as Moore's Law slows. Intel showed its first EMIB-plus-thick-core glass substrate sample (78×77mm, a 10-2-10 stack, 800 micron thick) at NEPCON Japan in January, reporting no micro-cracking as a mass-production readiness signal — though Intel's original 2023 target for commercialization was "by the end of the decade," so this is still a sample, not a product. Estimates of the glass-substrate market's size diverge by close to an order of magnitude between research vendors, so specific dollar figures here are not reliable enough to repeat as fact. As for 2D channel materials like graphene and molybdenum disulfide, the EU-funded 2D-EPL pilot line ran several multi-project-wafer runs in 2022 and 2023, and a January 2026 Nature Communications paper demonstrated stretchable monolayer MoS2 transistors — but no authoritative source describes an industrial production timeline for 2D-channel devices. This remains long-horizon basic research, not something landing within the next five years.

Behind all of it sits real industrial policy money. Japan's Rapidus received ¥267.6 billion (about $1.7 billion) in government funding this February, followed by a further ¥631.5 billion (about $4 billion) approved by Japan's Ministry of Economy, Trade and Industry in April targeting 2nm production by 2027, then another ¥150 billion round in June — cumulative government support now runs to roughly ¥2.35 trillion. The EU published its first proposal for a "Chips Act 2.0" on June 3, building on an original Chips Act (in force since September 2023) that has already mobilized more than €52 billion in investment and around 46,000 jobs; the new version extends state aid to "first-of-a-kind" projects across the value chain, from raw materials through packaging, with a stated goal of doubling the EU's global semiconductor market share to 20% by 2030.

2. PFAS Alternatives: Europe, the US and Japan Are Visibly Diverging

This is the direction with the most genuine disagreement between regulators and industry, and it's worth reporting the disagreement rather than smoothing it over. At the EU level, a joint restriction proposal from Denmark, Germany, the Netherlands, Norway and Sweden covers roughly 10,000 PFAS substances across about 14 sectors. ECHA's Committee for Risk Assessment adopted its final opinion and the Committee for Socio-Economic Analysis published a draft opinion in March, with public consultation running through May 25; a final opinion is expected by year-end. But even after that, a formal Commission legislative proposal isn't expected before 2027, and legal analysts don't expect actual adoption before Q3 2027, with application unlikely to start before 2029 — which contradicts a lot of industry material still circulating a "2027 compliance" narrative. Worth noting: the exemptions are expanding, not shrinking. ECHA widened derogated uses from 26 to 74 categories; semiconductor manufacturing has a proposed 13.5-year derogation (an 18-month general transition plus a 12-year extension), and medical devices and "green energy technologies" could get exemptions of up to 12 years. Separately, a broader "REACH 2.0" reform (unrelated to PFAS specifically) was shelved this year after industry pushback — the German Chemical Industry Association's (VCI) president Wolfgang Große Entrup put it as "competitiveness needs room to breathe, not another regulatory onslaught."

The US picture is messier. The EPA has repeatedly pushed back the start of TSCA 8(a)(7) PFAS reporting, now set for January 2027 or 60 days after a revised final rule, whichever comes first, while it works through thousands of comments on its November 2025 proposed revisions. On drinking water, the EPA is keeping the 4 parts-per-trillion limits for PFOA and PFOS but plans to extend compliance deadlines to 2031, while moving to rescind the limits set for PFHxS, PFNA, GenX and the PFBS-inclusive hazard index — a partial rollback that environmental groups have openly contested. 3M completed its own PFAS manufacturing exit on schedule at the end of 2025, as it committed to in 2022, though it's still working through third-party PFAS-containing components (lithium-ion batteries, circuit boards, seals) in its supply chain, and customer-transition work continues into 2026.

On the substitution side, individual company moves are worth naming rather than generalizing: Solvay is exiting trifluoroacetic acid and its derivatives globally, including at its Bad Wimpfen site in Germany, citing structural demand decline — while investing in a non-TFA brazing-flux product line at that same site. BASF plans to exit its PFAS business by 2028. Daikin is running the opposite play in a sense — it developed a new fluoroelastomer process that eliminates or reduces surfactants, with product sales starting this year, while pushing emulsifier recovery at its US and EU plants to 99% and setting a new 99.9% wastewater PFAS-capture target — but the company's own timeline for transitioning its broader fluoropolymer line is "around 2030." AGC is investing roughly ¥15 billion in a Kitakyushu production line for its surfactant-free FORBLUE S-series ion-exchange membrane, targeting a June 2026 production start and ¥30 billion in sales by fiscal 2030. Fluorine-free firefighting foams using hydrocarbon surfactants are already commercially available, but peer-reviewed literature is explicit about the trade-off — slower extinguishment than traditional AFFF on large fuel fires, and some hydrocarbon surfactants showing higher acute aquatic toxicity.

Japan's regulatory path is visibly different from Europe's. Rather than a blanket restriction, Japan is tightening its drinking-water standard: PFOS and PFOA are moving from a "provisional target value" (50 nanograms per liter combined) to a binding standard under the Water Supply Act starting this fiscal year, with mandatory quarterly testing for utilities. Only PFOS, PFOA and PFHxS are currently regulated substances in Japan, aligned to the Stockholm Convention — far narrower than the EU's roughly 10,000-substance proposal. AGC's CEO Yoshinori Hirai said plainly in early 2024 that the EU proposal wouldn't apply uniformly to every substance, only specific ones, and that AGC would keep tracking regional regulatory divergence. Daikin frames fluorine materials outside the Stockholm Convention list as indispensable to carbon-neutral transition, semiconductor, automotive and energy applications — even as it invests in substitution research.

Germany and the EU's "essential use" pushback is worth recording as genuine industry-versus-NGO conflict, not one-sided PR. VCI's core position is that a blanket ban would prohibit "several thousand substances with vastly different properties" without individual risk assessment, arguing for a differentiated approach that keeps semiconductor manufacturing, energy and climate technology, medical technology and safe industrial operations exempted — without itemizing which consumer-facing uses (textiles, cosmetics) it concedes do have alternatives. In January 2025, NGO coalitions including ChemSec's Forever Lobbying Project, Germany's BUND and PAN Germany publicly accused VCI of lobbying to protect commercial interests, disputing the "no alternatives" framing and arguing substitutes are more available than industry claims. ECHA's own committees seem to have partly sided with the differentiated approach — the expansion from 26 to 74 derogated uses, the multi-year exemptions for semiconductors, medical devices and green-energy tech — while still proposing to restrict roughly 10,000 substances overall; consumer categories like textiles, cosmetics and food-contact packaging are, per German trade coverage, on a faster phase-out track starting October 2026.

3. Green Methanol: The Regulatory Backstop Just Slipped a Year, and Maersk's Ships Are Still Mostly Running on Conventional Fuel

The first thing worth correcting here is a widely repeated assumption: a lot of industry material still treats 2027 as the settled start date for shipping's fuel-standard compliance. In October, at a second extraordinary session of the IMO's Marine Environment Protection Committee, a 57-49 vote (led by Saudi Arabia, under heavy US pressure) postponed adoption of the Net-Zero Framework by a full year; talks resume in October 2026. The regulatory driver that was supposed to force methanol demand is currently in limbo.

Maersk is the most concrete buyer in this space. As of this February, it has roughly 20 methanol-dual-fuel vessels in operational service — six 17,480-TEU large ships completed by January, plus the first of a new six-vessel 9,000-TEU mid-size class delivered three months early in February, with the rest arriving through 2026 and into early 2027. But supply is visibly lagging demand: Maersk's contracts with Goldwind (500,000 tonnes/year of bio/e-methanol, signed November 2023) and LONGi (bio-methanol from agricultural residues, signed October 2024) together cover only just over half of what the 2027 fleet will need. Maersk's own chief shipbuilder has acknowledged the company knew from the start it couldn't run these ships on green methanol at delivery — most of the fleet is currently burning conventional fuel while green supply slowly ramps up.

Actual operating e-methanol plants are still countable on one hand. The Kassø facility in Denmark, built by European Energy and Mitsui, is the world's first large-scale commercial e-methanol plant — inaugurated in May with roughly 42,000 tonnes/year of capacity, offtake agreements with Maersk, LEGO and Novo Nordisk, more than €1 billion invested, a €228 million award from Germany's Hydrogen Bank for an electrolysis expansion, and a green-financing deal with Mitsui signed this January. China's picture mixes real construction with paper announcements. On the real side: Shanghai's first domestically built green methanol plant (100,000 tonnes/year, running on biogas and food waste) broke ground in January and began commercial operations in December; Geely's project in Alxa, Inner Mongolia targets 500,000 tonnes eventually, with a 100,000-tonne first phase; a biomass project in Kangping, Liaoning has a 100,000-tonne-per-year first phase starting construction in 2026; industry estimates put total under-construction capacity nationally above 3 million tonnes. On the paper side: a widely publicized "world's largest," 1-million-tonne-plus green methanol megaproject in the same Alxa zone — tied to Geely founder Li Shufu's business network, with a reported ¥18.53 billion investment — was quietly cancelled on June 16 with no public explanation. That's a concrete example of the gap between announced and actual capacity.

Japan and Germany are, for now, mostly at the intent and technology-licensing stage. Idemitsu Kosan and Mitsubishi Gas Chemical announced marine methanol supply cooperation last October — shared tanks and bunkering vessels, synthetic plus bio-methanol — with supply starting fiscal 2026, but volumes explicitly undecided in the announcement. In February, a five-party partnership (Mitsui O.S.K. Lines, Mitsubishi Gas Chemical, the city of Yokohama, Kokuka Sangyo and Idemitsu) completed Japan's first ship-to-ship methanol bunkering. In Germany, C1 Green Chemicals produced its first batch of methanol via a novel low-CO2 process at the Leuna chemical park last September — local press called it a possible commercial breakthrough, but it's still pre-scale. Thyssenkrupp Uhde is mainly in the technology-licensing business, selling its "green methanol" power-to-methanol process to customers like Finland's Koppö Energia, and working with BASF, OMV and the German Aerospace Center on methanol-to-sustainable-aviation-fuel technology.

On cost, mature-technology e-methanol currently runs about $460 per tonne to produce, versus roughly $230 for fossil methanol — about double — with the general methanol market averaging around €700 per tonne in 2025. The IEA puts current renewable methanol production at under 200,000 tonnes a year, against a roughly 120-million-tonne total global methanol market — effectively a rounding error — constrained mainly by the cost of green hydrogen and captured CO2 feedstock, with costs not expected to fall to the $250–630 range until 2050. In plain terms: the number of real, revenue-generating e-methanol plants remains a handful, with Kassø as the flagship example.

4. Advanced Polymers: Carbon Fiber Is Genuinely Scaling, Recyclable Thermosets Are Still at Pilot Stage, Self-Healing Materials Are Mostly Lab Work

This is the direction where maturity levels vary the most, and it's worth being explicit about what's commercial versus what's still pilot- or lab-stage, since this space has more than its share of overstated claims.

Carbon fiber is the most solid part of the story. Toray Group is expanding capacity more than 20% to 35,000 tonnes/year starting in 2025 — an extra 3,000 tonnes at its Spartanburg, South Carolina site, new lines at Gumi in South Korea, its Abidos plant in France going from 5,000 to 6,000 tonnes in 2025, and a further European facility due fully online in the second half of 2026 — driven mainly by pressure-vessel demand for hydrogen and CNG tanks. Teijin launched a new carbon-fiber brand, "Tenax Next," in March, with two products already for sale at JEC World 2025: a filament yarn using ISCC PLUS mass-balance bio-based and recycled feedstock with about 35% lower manufacturing CO2 than the conventional grade, matching mechanical properties for drop-in qualification in aerospace and automotive; and a chopped fiber recycled from production offcuts carrying a QR-code-based digital product passport. These are commercial products, not pilot samples. High-performance thermoplastics like PEEK, PPS and PPA are also expanding into EV battery components: Victrex, which holds around 40% of the global PEEK market, has invested more than $150 million in Asian capacity since 2021 for EV demand, and Evonik's flame-resistant VESTAKEEP PEEK is aimed at battery-module and motor insulation, backed by a dedicated PEEK magnet-wire lab in Shanghai — though these capacity and share figures come from trade and market-analysis sites rather than primary company filings, so treat them as directional.

Chemically recyclable thermosets are a genuine research direction, but the accurate label right now is pilot, not commercial. Covestro announced a pilot-plant investment in Leverkusen, Germany last December for chemically recycling its Vulkollan polyurethane elastomers — breaking the material back down into chemical building blocks, with more than 90% mass recovery and up to two-thirds lower carbon footprint versus virgin material. Covestro's own framing is that this is "beyond lab scale," with roughly a year to technical completion and investment in the double-digit millions of euros — a bridge toward commercial scale, not commercial scale itself. Market-research estimates put the broader "recyclable thermoset resins" category (vitrimers, cleavable epoxies) at about $1.8 billion and 450,000 tonnes a year in 2025, naming Covestro, BASF, Hexion, Huntsman and Mitsubishi Chemical as participants — but that's an aggregated market-report figure, not verified per-company output, so it should be read as directional only. Germany's Fraunhofer Institute for Applied Polymer Research and the IKV plastics-processing institute in Aachen both published explainer pieces last year on vitrimers as a route to recyclable carbon-fiber composites — still applied-research framing, with no named commercial production line.

On bio-based and biodegradable polymers, global PHA (polyhydroxyalkanoate) production capacity reached about 115,000 tonnes/year in 2025, with Danimer Scientific, RWDC Industries and Kaneka as the main producers — still small next to PLA and commodity plastics, and priced three to four times polypropylene. Total global bioplastics capacity is projected to roughly double, from 2.31 million tonnes in 2025 to 4.69 million tonnes by 2030, with PHA's share of that rising from about 4.7% to 16.8% globally (2.9% to 9.5% in Europe specifically). The EU's Packaging and Packaging Waste Regulation (2025/40) was published in the Official Journal on January 22, 2025, entered into force on February 12, and applies starting August 12, 2026 after an 18-month transition, with recycled-content and digital-labelling obligations staggered further into 2027 and beyond. Self-healing and stimuli-responsive polymers remain, by contrast, overwhelmingly lab-stage work — active 2025 peer-reviewed research on polyimine vitrimers and dynamic networks, but nothing reporting industrial deployment. The one concrete commercial claim found (a capsule-based self-healing radome coating, said to be FAA-approved for commercial aircraft) traces to a niche coatings vendor in 2023 and appears only in market-research summaries, without independent confirmation from a primary aerospace or chemicals company release — so any prediction of "self-healing aircraft parts by 2028" should be read as market-report speculation rather than a verified roadmap.

On policy and recycling capacity, Germany's federal research ministry funds plastics-recycling-technology scale-up through its KuRT program, and a joint industry position paper from the country's plastics and waste-management associations disclosed that Germany's average recycled-content share in plastic products currently sits at just 15%, with PET bottle recycled-content quotas already in force since 2025, all-packaging quotas from 2030, and mandatory effective collection from 2035 — while roughly 7 million tonnes a year of plastic from municipal waste is still landfilled across Europe. The paper explicitly backs chemical recycling as a complement to mechanical recycling for mixed, composite and contaminated waste streams. China's chemical-recycling capacity is still small relative to mechanical recycling (about 1% versus 99% of the domestic recycling market) but is scaling: Dongyue Chemical's 200,000-tonne-per-year pyrolysis unit in Jieyang, Guangdong — China's largest single chemical-recycling unit — came online in the second quarter of 2025 but was shut down for upgrades, with restart expected in January 2026; Shandong Huicheng has announced plans for 5.4 million tonnes a year of capacity nationwide. China is expected to account for more than 35% of global recycled-plastic output in 2025.

5. AI-Designed Catalysts: Prediction Is Advancing Fast, but There's a Real Gap Between "Made in the Lab" and "Running in a Plant"

This is the direction most prone to overstatement, and the most important distinction to hold onto is between "AI predicted a candidate material" and "an AI-designed catalyst is running in industrial production" — these are very different claims.

Google DeepMind's GNoME project, published in Nature in November 2023, predicted 2.2 million candidate crystal structures, 380,000 of which were flagged as stable enough to be worth synthesizing; DeepMind reported that 736 had been independently created by outside labs, and a companion paper from Lawrence Berkeley National Laboratory's "A-Lab" reported 41 new materials synthesized autonomously. The result drew fast, pointed pushback. Seven researchers from University College London and Princeton, led by UCL's Robert Palgrave, posted a non-peer-reviewed critique on ChemRxiv in January 2024 concluding that A-Lab "didn't manage to make a single novel inorganic material" — most of the 35 claimed new compounds were, in their reading, misclassified mixtures of known compounds, with the AI's interpretation of diffraction patterns performing worse than a novice human researcher's. Gerbrand Ceder of UC Berkeley, the A-Lab paper's lead author, dismissed the critique as unreviewed and error-ridden. Separately, UC Santa Barbara's Anthony Cheetham and Ram Seshadri argued in Chemistry of Materials that GNoME's outputs lack the combined "novelty, credibility, and utility" needed to call them real materials discoveries, calling them "compounds" rather than functionally validated materials. A 2025 Chemical & Engineering News piece flagged duplicate and near-duplicate structures as a recurring problem across crystallography databases more broadly.

Microsoft's MatterGen, published in Nature in January 2025, takes a different approach — a diffusion model that generates candidate structures directly rather than screening an existing library, trained on 608,000 known stable materials. Microsoft says its outputs are more than twice as likely to be both novel and stable as prior methods, and land 15 times closer to the local energy minimum. There's a reasonably solid validation case here: a collaboration with the Shenzhen Institutes of Advanced Technology synthesized a MatterGen-designed material, TaCr2O6, matching the prediction. Microsoft's Azure Quantum Elements team separately screened 32 million candidate battery materials down to a novel solid-state electrolyte, which the Pacific Northwest National Laboratory synthesized and tested end to end, using notably less lithium than comparable materials. Meta's Open Catalyst Project (OC20/OC22) established the standard benchmark for machine-learning interatomic potentials in heterogeneous catalysis — ammonia synthesis, hydrogen production, CO2 utilization — aiming to replace costly density-functional-theory calculations in high-throughput screening. Its 2024 follow-up, OCx24, is the first phase to systematically bridge computation and experiment, synthesizing and characterizing 572 samples — itself an acknowledgment that the earlier phases stayed purely computational. No chemical company was found publicly attributing production catalyst screening to OC20/22; adoption so far is at the research-tooling level, not industrial deployment.

Japan's progress is concentrated in refining the computational methods themselves. Science Tokyo (formerly Tokyo Institute of Technology) published a method in npj Computational Materials this April combining a conditional variational autoencoder with a neural-network potential to design platinum-alloy fuel-cell catalysts — over six iterations, mean overpotential dropped from 1.126 to 0.520 volts, alloy formation energy improved, and the share of viable candidates rose from zero to 19%, with platinum-yttrium alloys (especially near a Pt3Y ratio) identified as most promising. This remains computational; there's no reported experimental synthesis yet. Science Tokyo and Kyushu University also open-sourced a generative-AI catalyst-design framework called CatDRX last October, and University of Tokyo professor Teruyasu Mizoguchi's "MatAgent" — a large-language-model-orchestrated multi-agent system for autonomous crystal design that emphasizes interpretability over black-box generation — was reported by Nikkei in December and published in Cell Reports Physical Science. On the industrial side, Nikkei reported chemical maker Nippon Shokubai piloting generative AI with startup Stockmark to raise R&D hit rates, and earlier Nikkei coverage noted Sumitomo Chemical cutting materials-property-prediction time roughly tenfold using AI.

Germany has the most complete computation-to-experiment loop found in this research. The Fritz Haber Institute's theory department, led by Karsten Reuter, working with BASF through their joint BasCat/UniCat lab, used a self-driving lab platform to search roughly 10^13 possible promoter combinations for propane-to-propylene conversion, needing fewer than 50 experiments to find a multi-promoter catalyst matching the performance of decades of empirical industrial work. The team specifically emphasized that the system could explain, not just predict, why the catalyst worked — a "gray-box" rather than black-box result, published in ACS Catalysis this March with a companion report from the Max Planck Society in April. BASF separately runs an "AI reactor" that plans, executes and analyzes experiments roughly 20 times faster than manual workflows, and maintains a standing machine-learning and explainable-AI catalyst research collaboration with TU Berlin. On the policy side, Germany's federal research ministry launched the Mat2Twin program and the "MaterialVital" materials-hub initiative last year, framing AI and digital-twin methods as tools for predicting material properties — funding-level support rather than a specific catalyst deliverable.

Taken together, a 2026 review in Angewandte Chemie names the field's central problem plainly: a persistent "complexity gap" between what generative and machine-learning models predict and what can actually be synthesized or measured, compounded by a "credibility gap" — AI models optimize for pattern-consistency rather than empirical truth, sometimes producing plausible-looking but false results, and lack the instinctive skepticism a human researcher applies to an implausible one. By 2026, the more honest framing in the field has shifted from "can we test this candidate" to "which candidate should we even test" — triage, not synthesis capacity, is the actual bottleneck. Across every case above, the pattern holds: computational prediction and narrow experimental proof-of-concept — single compounds, lab scale — are moving fast, but no source reviewed here documents an AI-originated catalyst that has actually replaced a deployed industrial production catalyst. The Fritz Haber/BASF and Science Tokyo work come closest to industrial relevance, but both are described as matching or approaching existing performance, not displacing it.

---

References

Semiconductor materials

• NEDO — ROHM SiC Green Innovation Fund milestone
• Infineon — Kulim, Malaysia 200mm SiC fab opening
• imec — High-NA EUV lithography
• ZEISS SMT — High-NA EUV optics
• Nikkei xTech — photoresist chemistry shift (CAR to MOR)
• Merck KGaA / EMD Group — semiconductor materials
• TrendForce — Intel glass-substrate sample
• SEMI — 2026/2027 equipment spending forecast
• Nature Communications — stretchable monolayer MoS2 transistors
• Rapidus — funding announcement
• European Commission — European Chips Act policy page
• Germany's BMFTR — Mikroelektronik-Strategie der Bundesregierung

PFAS alternatives

• ECHA — PFAS draft opinion consultation, spring 2026
• White & Case — Europe's PFAS restriction proposal
• C&EN/ACS — REACH revision shelved
• US Federal Register — TSCA 8(a)(7) reporting rule modification
• Star Tribune — 3M ends PFAS manufacturing on schedule
• ACS Omega — fluorine-free firefighting foams
• Solvay — press release on TFA exit and Germany investment
• Daikin Chemicals — PFAS sustainability page (Japan)
• Nikkei xTech — AGC CEO on EU PFAS scope
• Nikkei — Japan drinking-water PFAS standard
• VCI — "PFAS: Pauschalverbot ist keine Lösung"
• ChemTrust — NGO coalition statement on VCI lobbying

Green methanol

• IMO — net-zero shipping talks to resume in 2026
• DNV — IMO Net-Zero Framework decision delayed
• Maersk — new mid-size methanol vessel class deployed
• Maersk — long-term methanol sourcing deal (LONGi)
• European Energy — Kassø e-methanol facility inaugurated
• ccement.com — 100-million-tonne green methanol project cancelled
• Mitsubishi Gas Chemical — marine methanol supply cooperation
• Handelsblatt — C1 Green Chemicals methanol breakthrough
• IEA — World Energy Outlook 2025

Advanced polymers

• Covestro — pilot plant for chemical recycling of elastomers
• CompositesWorld — Toray carbon fiber capacity expansion
• Teijin — "Tenax Next" carbon fiber brand launch (Japan, primary PDF)
• European Bioplastics e.V. — market data
• Gleiss Lutz — EU Packaging and Packaging Waste Regulation briefing
• German plastics-industry associations — joint circular-economy position paper (primary PDF)
• Argus Media — recycled polymers market 2025

AI-designed catalysts

• DeepMind — millions of new materials discovered with deep learning
• The Register — "novel" AI-made materials not actually new
• Chemistry of Materials — is AI driving materials discovery?
• Microsoft Research — MatterGen generative materials design
• Meta — Open Catalyst Project
• arXiv — Open Catalyst Experiments 2024 (OCx24)
• Science Tokyo — AI-designed fuel-cell catalyst candidates
• Max Planck Society — explainable AI catalyst discovery with BASF
• Angewandte Chemie — accelerating catalyst discovery with large AI models