NH Nonghyup Insurance is the only non-life insurance company to invest in "Revelion," an artificial

NH Nonghyup Insurance will invest 10 billion won equity in AI semiconductor designer Revelion, which reached a ~2 trillion won valuation after merging with Sapion Korea. The insurer cites Revelion's REBEL chip as having superior power efficiency and price competitiveness versus peers like Nvidia and is positioning to capture global data-center growth by expanding alternative high‑tech investments. NH Nonghyup Financial Group also launched a 1 trillion won "NH Korea Win-Win Growth Fund" to be fully invested by pan‑agricultural affiliates, signaling a strategic shift toward venture and growth tech allocations.

Analysis

Domestic institutional capital being steered into local AI fabless design materially shortens the funding runway for chip startups and forces incumbent ecosystems to internalize a new competitive axis: national procurement and engineering partnerships rather than just pure price/performance. That shifts value toward foundries and advanced packaging partners that can turn designs into shippable silicon on a 12–36 month cadence; expect firms with flexible multi-node capacity to capture outsized incremental revenue if several domestic designs progress from tape-out to production in parallel.

Near-term upside for headline valuations is conditional and binary: public market re-rating hinges on concrete datacenter design wins and third-party benchmarks — milestones that typically arrive on 6–24 month timelines. Countervailing risks include software-stack lock-in (multi-year switching costs at hyperscalers), yield/margin risk at scale, and potential US export controls that could choke access to critical IP or advanced nodes; any of these can reverse narratives quickly and pressure early-stage multiples.

Second-order winners are HBM and interconnect suppliers, and OSATs that reduce time-to-market via co-packaging services; second-order losers are OEMs who pay a premium for CUDA-optimized accelerators but face procurement pressure if validated alternatives undercut ASPs. From a macro standpoint, government-backed allocations increase the probability of preferential procurement and M&A interest from hyperscalers, compressing the window for incumbents to respond before structural share loss, but true displacement is likely measured in years, not months.

Our working hypothesis: this is an underwritten acceleration of the domestic ecosystem that raises medium-term competitive pressure on incumbents but does not immediately negate software/ecosystem moats. Positioning should therefore hedge against a 20–30% ASP compression over 12–24 months while capturing upside in foundries/packaging beneficiaries and selectively participating in private-market allocations where valuation asymmetry exists.

Key Decisions for Investors

• Pair trade (12–24 months): Long advanced-node foundry/packaging exposure (TSM, SSNLF) 1–2% NAV vs buy-protected short on NVDA — implement via buying NVDA 9–12 month puts at ~0.25 delta and sell farther OTM puts to form a put spread. Rationale: capture upside from increased wafer/order flow while capping option premium; payoff if incumbents face ASP pressure. Risk: NVDA’s software moat holds; max loss = net premium (~1–2% NAV), upside 3–6x if NVDA falls 20–30%.
• Direct equity long (12–18 months): Overweight Samsung (SSNLF) or TSM — buy 12-month call overwrites on a 1–2% NAV size if available, targeting 20–40% upside if local designs scale. Rationale: foundries/packagers capture manufacturing share irrespective of which design wins. Downside: global capex slowdown or foundry share loss; set stop-loss at 15% drawdown or hedge with index puts.
• Private/co-invest allocation (18–36 months): Commit to selective private funds or co-investment vehicles focused on domestic AI fabless and IP (size 0.5–1% NAV initially). Rationale: participate before public re-rating and secure negotiated terms; expect 3–7 year liquidity horizon. Risks: execution/milestone risk and follow-on capital; require staging by technical milestones (tape-out, HVM).
• Event trigger: If independent benchmarks demonstrate >20% perf/Watt advantage and design wins from hyperscalers appear, rotate 50% of NVDA hedge into specialist suppliers (HBM vendors, OSATs). Conversely, if no benchmarks in 12 months, take profits on foundry longs and redeploy into broader AI software ecosystem names.