China optical fibre industry: UBS sees China optical-fibre oversupply fears as overstated as AI data-center demand and supply bottlenecks support pricing.
UBS expects global AI data-center roll-outs to shift fibre demand from telecom-budget-driven to consumption-driven, keeping high-end fibre supply tight through at least 2027-28E. It reiterates Buy and top-pick status on YOFC-H, while maintaining Neutral on YOFC-A.
Summary
UBS expects global AI data-center roll-outs to shift fibre demand from telecom-budget-driven to consumption-driven, keeping high-end fibre supply tight through at least 2027-28E. It reiterates Buy and top-pick status on YOFC-H, while maintaining Neutral on YOFC-A.
- The sector corrected 37% from its June peak amid peak-earnings and capacity-expansion concerns.
- UBS forecasts global fibre demand growth of about 11% CAGR in 2026-30E, with data centers contributing more than 45% of demand by 2030E.
- Data-center fibre demand is forecast to grow at a 37% CAGR in 2026-30E.
- Preform bottlenecks, equipment lead times and customer qualification should delay effective new supply.
- UBS expects fibre prices to keep rising through 2027E, though at a slower pace.
- YOFC-H trades at a 57% discount to YOFC-A; UBS expects the discount to narrow.
Report Interpretation
Overview
UBS examines whether announced capacity expansion will recreate an optical-fibre oversupply cycle in China. Its conclusion is that global AI data-center demand, a healthier domestic procurement mix and substantial barriers to effective high-end supply should sustain pricing and favor established manufacturers, particularly YOFC-H.
Core views
UBS argues that the market has overestimated the likelihood that fibre-capacity expansion will trigger a repeat of prior China oversupply cycles. The sector has corrected 37% since its June peak because investors fear a sharp fibre-price decline and peak earnings around 2027E. UBS instead sees demand becoming less dependent on domestic telecom capital-expenditure budgets and increasingly tied to global AI data-center construction, which it believes is more closely linked to actual infrastructure deployment and therefore less prone to speculative inventory build-up. The demand model is central to the thesis. UBS forecasts global optical-fibre demand to grow about 11% annually in 2026-30E, led by 37% CAGR data-center demand while telecom demand grows 5%. Data centers are projected to account for more than 45% of total fibre demand by 2030E, versus just over 10% in 2025. The report attributes demand growth to larger AI clusters, rising GPU and ASIC bandwidth, data-center interconnection and, from 2028 onward, optical connectivity moving further into racks and pods. It expects scale-out and scale-across deployments to drive growth through 2026-28, with scale-up becoming an additional driver from 2028. UBS sees evidence that China capacity is increasingly being absorbed by global demand. Chinese and overseas fibre prices have converged since late 2025, compared with China historically trading at a 20-40% discount. China fibre-cable export value has increased meaningfully since 2026, and global customers are increasingly using multi-year long-term supply agreements. UBS also notes that Chinese manufacturers account for more than 60% of global production capacity, positioning established suppliers to participate in overseas demand and to expand into higher-value data-center cabling and connectivity products. Domestic demand is also seen as improving in quality rather than simply volume. UBS expects China telecom capital-expenditure mix to shift from 65% telecom and 35% computing networks in 2026E to 50/50 in 2028E as AI and cloud investment rises. It argues that computing-network investment offers Chinese telcos better returns than traditional 4G/5G investment and should make procurement less focused on aggressive cost minimization. In 2026, postponed tenders ultimately resulted in China Mobile's latest fibre tender being finalized at only a 20% discount to spot prices, which UBS treats as evidence of a more rational pricing environment. On supply, UBS expects the market to remain tight through at least 2027-28E, with utilization above 90%. Existing manufacturers may optimize or expand brownfield facilities by late 2026 to early 2027, but such expansion is limited. Second-tier Chinese additions announced in 2Q26 are expected from late 2027 or 2028, while overseas additions are expected mainly from late 2028 onward. UBS estimates that greenfield projects require at least three years before effective supply emerges, especially for new entrants, because of equipment procurement, talent, clean-room construction, environmental approvals, process optimization and customer qualification. It therefore discounts announced capacity from newer players when estimating effective industry supply. The report emphasizes that preform manufacturing is the key bottleneck and a source of competitive differentiation. Leading Chinese producers have proprietary equipment, accumulated process knowledge, established customer relationships and more self-sufficient preform capacity. New entrants may initially add standard G.652.D capacity, but higher-end products require more specialized manufacturing and qualification. Demand is shifting toward G.657.A1/A2 for scale-out, G.654.E and hollow-core fibre for scale-across, and potentially polarization-maintaining fibre for scale-up. UBS expects this specification upgrade to keep high-end data-center fibre structurally tight even as conventional-fibre supply expands from 2028-29E, preserving stronger pricing power and profitability for technology leaders. Valuation concerns are a further focus. The sector rallied more than 300% to its June peak before the 37% correction, though it remained up 170% year to date. NTM consensus P/E de-rated from 40-50x to 23x, or 0.2 standard deviations above its historical average, while global peers including Corning and Prysmian remained more than one standard deviation above historical averages. UBS believes current China valuations already reflect oversupply fears. It expects upstream preform tightness and AI data-center demand to keep prices rising through 2027E, albeit more slowly, despite its base-case assumption that fibre-attached drone demand tapers from 2027E. For YOFC, UBS maintains Buy on YOFC-H with a HK$330 price target and Neutral on YOFC-A with a Rmb510 price target. YOFC-A trades at 34x NTM consensus P/E, a 79% premium to domestic peers' average and close to Corning's 39x, which UBS views as fair given YOFC's technology, product quality and higher optical-fibre/interconnect exposure. YOFC-H trades at a 57% discount to the A-share, compared with a 23% discount for a copper-interconnect AI-PCB peer and premiums for several dual-listed optical and semiconductor-interconnect peers. UBS argues that offshore investors are overly discounting industry oversupply risk and underappreciating YOFC's global AI supply-chain position, overseas data-center penetration potential and downstream connectivity opportunity.
Analysis framework
UBS combines a global supply-demand outlook, data-center architecture and fibre-specification analysis, assessment of capacity-build timelines and entry barriers, domestic telecom procurement trends, peer valuation comparisons and separate P/E-based valuation of YOFC's H and A shares.
Methodology notes
Global optical-fibre supply-demand balance
UBS forecasts demand growth by end market and compares it with the timing and effective scale of capacity additions to assess fibre pricing and utilization.
AI data-center build-outs flowing through to fibre, preform and connectivity demand
The report links GPU clusters, data-center architectures and telco investment to fibre specifications, supplier demand and pricing power.
P/E methodology for YOFC H and A shares
UBS values both share classes using P/E and compares their forward multiples and H-A discount with domestic and global peers.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- YOFC-H (6869.HK)UBS's top pick and preferred way to express the industry thesis.
- Strengths
- Technology lead, product quality, preform capability, AI-related fibre and interconnect product exposure, and potential global data-center penetration.
- Weaknesses
- H-share investors are discounting potential industry oversupply and peak earnings.
- Comparison
- Trades at a 57% discount to YOFC-A; UBS notes many dual-listed AI hardware peers trade at an H premium or a materially smaller discount.
- Risks
- Weak telco capex, slower AI data-center roll-outs, raw-material volatility and faster competitor capacity expansion.
- YOFC-A (601869.SH)Covered domestic share class of YOFC.
- Strengths
- UBS views its technology, product quality and optical-fibre/interconnect mix as supporting a premium to domestic peers.
- Weaknesses
- Already valued broadly in line with global leaders, limiting UBS's relative preference.
- Comparison
- Trades at 34x NTM consensus P/E, a 79% premium to domestic peers' average and close to Corning's 39x.
- Risks
- Weak telco capex, slower AI data-center roll-outs, raw-material volatility and faster competitor capacity expansion.
- Corning (GLW.N)Global comparable and evidence source for AI data-center fibre demand.
- Strengths
- Referenced as a global leader with major long-term agreements supporting fibre and connectivity demand.
- Comparison
- Trades at 39x NTM consensus P/E versus 34x for YOFC-A.
- Hengtong (600487.SS)Domestic industry comparable.
- Comparison
- Included in UBS's China optical-fibre sector comparison set.
- Fiberhome (600498.SS)Domestic industry comparable and announced capacity expander.
- Comparison
- Included in UBS's China optical-fibre sector comparison set.
Key data
- Global optical-fibre demand CAGR~11% in 2026-30EDriven by data centers and supported by telecom demand.
- Data-center fibre demand CAGR37% in 2026-30EThe report's main incremental demand driver.
- Data-center share of fibre demand>45% by 2030EUp from just over 10% in 2025.
- Industry utilization>90%UBS expects elevated utilization as meaningful new capacity cannot arrive quickly.
- China telecom/computing capex mix65/35% in 2026E to 50/50% in 2028EReflects a shift toward AI data-center and cloud investment.
- China Mobile tender discount to spot20%Latest tender was finalized at only a 20% discount to prevailing spot prices.
- Sector price correction37% since the June peakAttributed to oversupply and peak-earnings concerns.
- YOFC-H discount to YOFC-A57%UBS expects this H-A discount to narrow.
- YOFC-H price targetHK$330Buy maintained.
- YOFC-A price targetRmb510Neutral maintained.
Impact & implications
UBS expects AI-led demand, delayed effective supply and higher product specifications to support elevated fibre prices and margins, particularly for manufacturers with preform capacity, technology and qualified customer relationships. It believes valuation concern is most mispriced in YOFC-H because its H-share discount does not reflect its potential role in global AI data-center supply chains.
Risks
- Weaker-than-expected telecom capital expenditure could reduce fibre demand.
- AI data-center roll-outs could be slower than UBS expects.
- Raw-material cost volatility could pressure industry profitability.
- Competitors could expand capacity faster than expected.
- The report's base case assumes fibre-attached drone demand tapers from 2027E onward.