Figure AI Founder's New Startup Hark Is Latest to Plan Family of AI Devices

Brett Adcock’s new startup Hark is developing a family of AI devices for personal and home use, positioning them as distinct from existing handsets, wearables and smart glasses. The founder says the products aim to be so essential their absence would feel like “a day of lost information.” This is an early-stage private-venture product announcement and is unlikely to move public markets beyond sector-level investor interest in AI hardware.

Analysis

The next wave of consumer AI devices will bifurcate spend between high-throughput cloud training and low-power edge inference; the valuable second-order effect is a reallocation of capex within the semiconductor stack toward inference accelerators, sensor fusion, and power-management ICs rather than general-purpose CPUs. Expect outsized revenue tailwinds for companies that capture design wins for bespoke SoCs and camera/IMU stacks, and for cloud providers that monetize model updates and heavy lifting behind the scenes.

Supply-chain impacts will be non-linear: contract manufacturers and test-and-pack services will see longer design cycles and higher NRE per SKU, raising breakeven volumes and favoring suppliers with scale or cross-device OEM relationships. This raises a barrier to entry for hobbyist OEMs and increases the value of platform-level partners (chip vendors, cloud orchestration, and security subsystems), pushing consolidation in mid-tier suppliers over 12–36 months.

Key risks and reversal catalysts are software lock-in and unit economics. If large platforms (Apple/Google/Amazon) embed similar capabilities in phones/earables or tie devices into subscription services, new form factors face severe adoption friction; likewise, rapid model compression or LLM-offloading to the cloud could reduce on-device compute requirements, reversing hardware demand within 12–24 months. Regulatory/privacy backlash or battery/thermal constraints that materially degrade UX would also delay meaningful consumer adoption beyond a multi-year horizon.

Key Decisions for Investors

• Long NVDA (NVDA) LEAPs — buy 12–24 month call options sized 1–2% NAV: thesis is outsized demand for inference and training GPUs. Target 30–50% upside if device wins ramp; set tactical stop at 25–30% premium erosion or if quarterly guidance reduces data-center growth materially.
• Long Qualcomm (QCOM) equity or 9–12 month call spread — exposure to low-power SoCs for always-on AI plus modem/IP licensing. Position size 1–1.5% NAV; expected asymmetric payoff if multiple device classes adopt its reference designs. Cut if design-win cadence slips by two consecutive quarters.
• Pair trade: short AAPL / long QCOM (equal notional, 6–12 month horizon) — hedge exposure to platform lock-in: short Apple to capture downside from ecosystem fragmentation; long Qualcomm to capture share of new form-factor SoCs. Keep small (0.5–1% NAV each leg), stop-loss if Apple announces a major new bundled AI hardware/service within 90 days.
• Long Ambarella (AMBA) or similar edge-vision specialists via 9–18 month calls — high optionality on camera/vision compute content in new devices. Small allocation (0.5–1% NAV) as a volatility play; expect binary outcomes tied to specific design wins, take profits on 50%+ move or cut at 40% downside.