One material, two demand curves: AI wants its electrical properties, batteries want its mechanical talent
A material that has quietly worked in the battery industry for years has suddenly been thrust into the spotlight of the AI supply chain — by NVIDIA’s materials selection.
As the materials roadmap for NVIDIA’s Rubin Ultra platform moves forward, PTFE (polytetrafluoroethylene) has rapidly become one of the hottest topics in the PCB supply chain. Supply-chain reports from SemiAnalysis and other sources indicate that NVIDIA has listed PTFE as a test material for NVSwitch-related boards in Rubin Ultra. As channel speeds climb toward 224G and even 448G, conventional high-speed copper-clad laminates (CCL) face mounting pressure from signal loss — and PTFE, with its extremely low dielectric loss, has naturally emerged as a candidate material for next-generation high-speed PCBs.
The capital markets have already responded: on August 27-28, Haohua Technology and Adtech (Wote) shares hit their daily price limits for two consecutive sessions, while shares of Kente, Lianyungang Lianrui New Materials and other PTFE-chain companies rallied sharply.
This, of course, raises a pressing question for the battery industry: if AI keeps pushing up demand for high-grade PTFE, will battery makers racing to industrialize dry electrodes and all-solid-state batteries run out of material?
Reading guide: this article reorganizes the original reporting framework and adds key PTFE material parameters (dielectric properties, fibrillation mechanism, loading-level estimates). All parameters are typical values from public sources, intended as order-of-magnitude references. Approximately 2,500 words.
1. Getting to Know PTFE: A Material Profile
PTFE is anything but new. This fluoropolymer — polymerized from tetrafluoroethylene monomer and nicknamed the “king of plastics” — has its backbone almost completely enveloped by fluorine atoms, forming a shield that is nearly impossible to breach. The payoff is fourfold: extreme chemical resistance, very low surface energy, a very low coefficient of friction, and dielectric constant/loss values among the lowest in the entire polymer family.
For decades it has quietly served chemical anti-corrosion, seals, films, cables and high-frequency communications; consumers know it best as the coating on non-stick pans. Today, two very different demand curves have landed on this same material at the same time:
| PTFE Material Profile (typical values) | |
|---|---|
| Dielectric constant Dk (1 MHz to 10 GHz+) | ≈ 2.1 (among the lowest of any polymer; nearly flat vs. frequency) |
| Dissipation factor Df (tan δ) | ≈ 0.0002–0.0005 (an order of magnitude lower than M9-grade hydrocarbon resins) |
| Continuous service temperature | -200°C to 260°C (melting point 327°C) |
| Coefficient of friction | ≈ 0.04–0.1 (among the lowest of any solid material) |
| Chemical resistance | Resistant to virtually all acids, bases and solvents except molten alkali metals |
| Shear fibrillation | Forms micro-fibrillar networks under high-speed shear/compression (the physical basis of dry electrodes) |
| Key weaknesses | Cold flow/creep under sustained load; high sintering temperature makes melt processing difficult; reductive-instability issues at low-potential anodes |
Note: Dk/Df are typical values for the neat resin; values change once formulated into laminates with glass fabric and fillers. Refer to vendor datasheets for specific grades.
2. The AI Demand Curve: Why NVIDIA Wants It
As data-exchange rates inside AI servers climb, material properties now directly determine how far a signal can travel reliably. The physics is straightforward: dielectric loss scales roughly with frequency × √Dk × tan δ. The higher the data rate and the higher the frequency band, the faster loss pressure compounds.
At this year’s GTC, NVIDIA unveiled the Kyber rack architecture for Rubin Ultra: 144 GPUs are interconnected through large mid-planes and NVLink switches, and interconnects that once relied on high-speed copper cables are increasingly giving way to large-scale PCBs. The result is thicker boards, more layers, and far stricter signal-integrity requirements.
| High-speed CCL grade (industry convention) | Df (approx.) | Typical application |
|---|---|---|
| M6 grade | ~0.002 | Mainstream for 112G SerDes |
| M7/M8 grade | 0.0008–0.0015 | 224G switch-chip boards |
| M9 grade (low-loss hydrocarbon) | 0.0006–0.001 | 224G high-speed layers, hybrid laminates |
| PTFE system | 0.0002–0.0005 | 448G candidate: the only mature resin system that is simultaneously “double-low” |
Note: typical ranges reported in industry publications; values vary by vendor and test frequency. Order-of-magnitude reference only.
It is worth stressing that what the supply chain is testing is not a wholesale replacement of today’s M9 materials with PTFE. The more pragmatic approach under evaluation is hybrid lamination — combining PTFE cores with conventional prepreg to balance dielectric performance, mechanical strength and manufacturing yield (pure PTFE’s tendency to cold-flow/creep and its high sintering temperature are hurdles that CCL processing must work around). In other words, PTFE’s entry into Rubin Ultra is a high-probability candidate still under validation, not a confirmed, announced BOM.
The supply side is already moving
| Company | Publicly disclosed PTFE positioning |
|---|---|
| Haohua Technology | PTFE capacity of 51,000 t/yr; high-end electronic-grade PTFE ready for industrialization, with long-term partnerships established with mainstream CCL makers (July investor-relations disclosure) |
| Juhua (Quhua) Group | 28,000 t/yr capacity as of end-2025, plus 37,000 t/yr under construction; a 7,000 t/yr high-performance PTFE project in progress; pursuing electronic and semiconductor customers |
| Dongyue Group | Interim report cites AI, semiconductors and new energy driving a demand recovery in fluoropolymers: external sales of fluoropolymer segment ≈ RMB 2.14 billion (+10.35%), segment profit RMB 439 million (+69.21%) in H1 |
In short, AI has not yet generated massive absolute PTFE consumption — but it has already begun to rewrite how Chinese fluorochemical producers position their premium product lines.
3. The Battery Demand Curve: From “Reserve Material” to Production Line
Why batteries are interested in PTFE is a completely different logic from AI. Batteries don’t care about dielectric numbers — they want the material’s ability to draw out a fiber network under shear.
| Comparison | Wet process (conventional) | Dry process (PTFE system) |
|---|---|---|
| Binder | PVDF (requires NMP solvent) | PTFE (shear fibrillation to form network) |
| Solvent use | Large amounts of NMP; coating, drying and solvent recovery needed | Essentially solvent-free; drying energy saved |
| Film formation | Slurry coating → drying | Powder mixing → fibrillation → roll-pressed self-supporting film |
| Typical binder loading | ~2–5% (cathode) | Can be reduced to ~0.7% (per Qingyan Electronics) |
| Significance for solid-state batteries | Solvents can react adversely with sulfide electrolytes | Solvent-free: fewer side reactions, lower energy use |
The principle is not complicated: PTFE’s molecular weight can reach millions, and its chain segments are flexible. Subjected to high-speed shear or compression, it fibrillates, and the fibrils weave a continuous three-dimensional network between active-material particles — in effect “kneading” the powder into a free-standing electrode film. This trait becomes even more valuable in the all-solid-state era: for sulfide systems in particular, using less or no solvent both eliminates a large chunk of drying energy and avoids side reactions between solvent and solid electrolyte.
What changed in 2026: from lab to engineering
In the past, PTFE-based dry electrodes were mostly confined to labs and equipment development. The signal from 2026 is that materials, equipment and pilot lines are advancing on all three fronts simultaneously:
Qingyan Electronics: in April, it unveiled a PTFE binder developed specifically for dry electrodes and completed the full chain of “binder → mixing/homogenization → fibrillation → double-side film formation → mass-production line.” Chairman Wang Chen told Gaogong Li Dian in an interview that the company screened more than 20 binders globally before keeping PTFE as its core choice, with loading reduced to about 0.7%. Its 0.1 GWh fully automated pilot line is in operation, and a 5 GWh line is planned for H2 2026.
Dongyue Group: solid-state-battery dry electrodes have been added to the next-stage R&D roadmap for its high-end PTFE.
Put these pieces together and PTFE’s role in China’s battery industry is shifting: from a “potential binder” to a critical link in the full engineering chain of “specialty material — fibrillation equipment — continuous film formation — GWh-scale lines.”
4. Equipment Orders Speak First: Signals of Real Demand
When it comes to whether material demand is real, equipment orders usually tell the story first. Since the start of 2026, domestic dry-electrode equipment has visibly accelerated through customer validation and delivery:
- Honggong Technology: per its interim report, it has upgraded its twin-screw extrusion technology into dedicated dry-electrode equipment, using screw design to solve PTFE fibrillation and metal-wear problems;
- Qingyan Honggong (Qingyan × Honggong): focused on mixing/homogenization, dry grinding and twin-screw fibrillation equipment;
- Qingyan Nacol (Qingyan × Nacol): dry film-forming equipment is advancing toward 1,200 mm width, 60 m/min for cathodes and 80 m/min for anodes;
- Lead Intelligent: launched a new mass-production grade integrated cathode-anode dry coating solution;
- Yinghe Technology: its dry dispersion, fibrillation and film-forming equipment has been delivered to top-tier customers;
- Putailai: its dry equipment has passed shipment acceptance.
Still, meaningful distance remains to mass-scale ramp-up. In April, Huang Xiaowei, VP of Ganfeng Lithium Battery, publicly noted that dry electrodes have made clear progress on thicker anodes, but the cathode side has not yet achieved a breakthrough — the main obstacle to large-scale adoption of the technology. Qingyan Electronics, for its part, calls 2027 the “year one” of dry-electrode industrialization.
5. Can They Really Compete for Material? Two Ledgers — Volume and Structure
The volume ledger: no near-term squeeze
China’s total PTFE market is already on a hundred-thousand-tonne scale — the existing capacity held by Haohua, Dongyue and Juhua alone is considerable. By contrast, current PTFE consumption from AI servers and dry electrodes alike is a thousand-tonne-class variable.
Batteries are the clearest example. In dry-electrode formulations, PTFE typically accounts for 0.5% to a few percent of the solids. Using ~0.7% loading as a rough estimate: 1 GWh of battery capacity requires active materials on the order of a thousand tonnes, but PTFE usage is only a handful of tonnes. Even if tens of GWh of all-solid-state dry-electrode capacity comes online, PTFE demand would most likely stay in the hundreds-to-thousands of tonnes range.
So, at least through 2026-2027, the idea of AI “buying out” PTFE and leaving battery makers without supply is not realistic.
The structural ledger: the real mismatch to watch
What deserves closer attention is a subtler mismatch: neither side wants commodity-grade PTFE.
| PTFE grade | Key performance requirements | Where the barriers are |
|---|---|---|
| Commodity grade Pipes, seals, coatings | Basic chemical and thermal resistance | Cost and scale |
| AI electronic grade High-frequency CCL, PCB | Ultra-low Dk/Df, purity, dispersion, batch-to-batch consistency | Polymerization process control + long-term qualification by CCL makers |
| Battery grade (dry binder) Dry electrodes | Appropriate molecular weight, particle morphology, stable and controllable fibrillation | Resin polymerization + fibrillation process + battery-maker validation |
High-frequency PCB applications demand exacting purity and batch consistency — dielectric loss and dispersion must be flawless. Dry electrodes, in turn, require PTFE with stable, controllable molecular weight, particle morphology and fibrillation behavior. Ordinary capacity dedicated to pipes and seals cannot simply be switched over to premium grades. Juhua has pointed out the essence of the barrier when discussing high-end fluoropolymers: the real moat is not raw-material cost but production process, quality control and long qualification cycles — a deviation in any step from feedstock purification to process control can downgrade the entire product.
6. The Real Competition: Not Tonnage, but Grades
This is the crux of the new round of competition in the PTFE industry. In the past, the industry discussed PTFE by comparing tonnage capacity. Going forward, the more important question becomes: within those tens of thousands of tonnes of capacity, how much can actually and consistently produce AI-grade or battery-grade product?
Leading producers are simultaneously ramping up high-performance investment: Juhua holds 37,000 t/yr of capacity under construction; Haohua’s high-end electronic-grade PTFE has entered industrialization; Dongyue is redirecting raised capital toward ultra-high-purity PTFE for semiconductor and other premium applications. The supply side has not stopped expanding — what is changing is the structure of supply.
The two curves also sit at different points on the timeline:
- AI incremental demand will likely arrive earlier: Rubin Ultra’s switch trays, NVSwitches and the follow-on orthogonal backplane have already entered clear engineering validation, while most domestic all-solid-state batteries remain at pilot or small-batch stage;
- Batteries have the higher ceiling: once dry electrodes scale from a few GWh of validation to tens or even a hundred-plus GWh of capacity, PTFE demand could quickly cross into the thousands of tonnes.
There are also variables inside the battery route itself. In low-potential anode systems such as graphite and lithium metal, PTFE suffers from reductive stability issues; the industry is exploring composite binders, surface modification and novel non-fluorinated polymers as backup options. Its status as the “core binder” is not written in stone.
7. Conclusion
AI has rediscovered PTFE’s electrical value; batteries are amplifying its fibrillation ability.
For the PTFE industry, this may matter more than a few extra thousand tonnes of demand. A traditional fluoropolymer long priced by the tonne is entering an era of pricing by grade, process and certification.