I/O Fund (Beth Kindig)
Beth Kindig — lead tech analyst & co-founder; dubbed the "Queen of Nvidia" for AI-semi calls dating to 2018; runs the I/O Fund alongside David Kindig.
Internet-native AI-investing influencer who publishes a transparent, members-only model portfolio with real-time trade alerts and weekly research, ~95%+ allocated to AI stocks across the three pillars she names: networking, inference, and energy. Built a following on Seeking Alpha, Medium, and X around early, public Nvidia and AI-infrastructure calls. The book leans into picks-and-shovels names (power, optical, memory) rather than just the headline chip.
- Not a fund — A research, media or advisory business with no managed book to score.
- Self-reported — Headline returns are manager- or firm-reported and unaudited.
Top holdings
held up to ~15% — I/O Fund's 2026 top pick and best April performer; disclosed initial buys at $16.64 / $17.04 (Apr 2025), trade-alerted to members. AI-datacenter power layer.
core long — Long-standing signature position; her early NVDA calls are the basis of the "Queen of Nvidia" tag. Specific current weight not disclosed publicly.
long — Placed at the top of I/O Fund's 2026 AI chip-stock rankings; HBM/memory exposure to the AI buildout.
She cites an optical-networking stock up 620%+ since Nov and a 10% position up 130% YTD; specific ticker is behind the paywall — not disclosed here.
Datacenter optics / transceivers for AI networking
Recent moves
Q1 2026: published a 60+ page "Top 15 AI Stocks for Q1 2026" report emphasizing networking, inference and energy; reiterated Bloom Energy as 2026 top pick and Micron atop AI-chip rankings (Medium/X, Jan–Apr 2026). In May 2026 flagged that semiconductor strength "may signal a market top" and trimmed her Nvidia allocation while keeping the long-term thesis (Seeking Alpha, May 2026).
Our take
One of the few genuinely public AI-investing books run by a named figure with a documented multi-year call history (NVDA from 2018), and the most credible of the viral cohort because the model portfolio is transparent to members with timestamped trade alerts. The honest limits: the headline 326%/29% returns are firm-reported and only loosely described as "audited" (no named auditor or SEC ADV-grade track on the public pages), current weights for the best performers sit behind the paywall, and the business model (paid research) rewards bullish, high-variance picks. Useful as a real-money signal on the AI power/optical/memory periphery, not as an independently verified return stream.
The audit is the real differentiator and stronger than a newsletter track record: an independent San Francisco accounting firm reconciles actual brokerage + blockchain (equity + crypto) statements to the reported track — reported annually, ~7 audits at ~$32.5k total, verified clean per the firm, though the exact covered period and auditor name aren't on public pages — so the trades and arithmetic are credibly real. What the audit does NOT establish is comparability to the '8th best hedge fund' it's benchmarked against — this is a gross, fee-free personal account, not net-of-2/20 outside capital, and a member's captured return (alert latency, slippage on thin optical/fuel-cell names, unseen position sizing) is not the model portfolio's return. The business is paid bullish research, so revenue scales with bold, concentrated, high-variance calls — though the May 2026 NVDA trim cuts against a pure-permabull read.
Thesis
The I/O Fund isn't a fund — it's Beth Kindig's audited, members-only model portfolio (a paid-research product), ~95% allocated to AI across her 'networking, inference, energy' pillars. The 326%/29.2%-annualized headline is genuinely brokerage-audited, but it's a fee-free single proprietary account marketed against net-of-fee hedge funds — a comparison the branding leans on harder than the math supports.
Concentrated long-only AI book run down the picks-and-shovels chain — power (Bloom Energy, her top 2026 pick), memory (Micron), optical (Lumentum) and NVDA — with real-time trade alerts to subscribers. Actively traded, not buy-and-hold: she trimmed Nvidia in May 2026 on a 'semiconductor strength may signal a market top' read while keeping the long-term thesis.
Assessment
- Rare transparency for the retail-research category: an audited real-money account with timestamped trade alerts, not a curated backtest.
- Documented multi-year public call history (NVDA from 2018) anchors a real, checkable reputation.
- Down-the-value-chain positioning (power/optical/memory) captured the 'second-derivative' AI trade, not just NVDA beta.
- Willing to trim and hedge (cut Nvidia May 2026 on a market-top signal) rather than run a static permabull book.
- 29.2% annualized is a gross, fee-free single-account figure benchmarked against net-of-fee hedge funds — not apples-to-apples.
- Subscriber returns ≠ portfolio returns: alert latency + slippage on illiquid optical/fuel-cell names erode the capturable edge.
- Paid-research model structurally rewards bullish, concentrated, high-variance calls and confident framing.
- ~95% AI concentration is the same AI-capex-cycle beta as the rest of the cohort, presented as stock selection.
- Key-person risk is total — the entire product is Beth Kindig.
- 2026-07In July 2026 the down-the-value-chain tilt was the damage rather than the diversifier: Bloom Energy -43.3%, Micron -22 to -27%, Lumentum -16.6%, while Nvidia held at +0.3% and Broadcom +3.1%.
- 2026-07The May 2026 Nvidia trim is harder to read as vindicated risk management: Nvidia was the one name that held up in July (+0.3%), while the untrimmed power, memory and optical book fell 17-45%.
- 2026-07July 2026 shows the exposure is narrower than generic AI-capex beta: the S&P 500 was -0.1% and the accelerator names flat-to-up while memory, power and optical fell 20-45%. Sub-sector risk, not index beta.
Record
Firm-reported 326% cumulative since May 2020 (~29.2% annualized) and +37% in 2025, framed as equivalent to the 8th best U.S. hedge fund — figures carried through BusinessWire/Morningstar (Feb 2026) and independently audited to brokerage statements, so the trades and math are credibly real. The open attribution question is skill vs. an unconstrained, fee-free, ~95%-AI account riding the biggest tech bull run in a generation: on a beta-adjusted, fee-adjusted, subscriber-capturable basis the 'beats Wall Street' claim is unproven, not disproven. Headline 2026 YTD figures are reported publicly (a strong AI-driven YTD as of spring 2026); current position weights and full-book detail sit behind the paywall.
- 2026-08The last published 2026 figure remains +48% through April 2026, with nothing posted since. It was earned largely by Bloom Energy (~+109% in April) and the optical and memory names that then fell 17-45% in July.
Risks & fit
- An AI-capex-cycle drawdown hits a ~95%-concentrated book with no diversification cushion.
- Bloom Energy and small optical names are volatile, momentum-driven and thinly traded — real slippage and reversal risk.
- Marketing-vs-reality gap: if the hedge-fund benchmark comparison is challenged, brand/subscriber-trust exposure.
- Key-person: no Beth Kindig, no product or track record continuity.
- A regime rotation back from picks-and-shovels to headline names would wrong-foot the three-pillar thesis.
A GIPS-style, independently-computed track showing that (a) subscriber-capturable returns net of realistic slippage are roughly in line with the 29.2% headline and (b) the figure is time-weighted and fee-adjusted comparably to the hedge funds it's benchmarked against would refute our gross-vs-net / category-error skepticism and upgrade the record to genuinely peer-comparable skill. Conversely, a full-cycle stretch where the concentrated AI book underperforms a simple AI-heavy index net of the subscription cost would confirm the beta-not-alpha read.
Analytically most useful as a transparent, audited real-money read on the AI power/optical/memory periphery — a signal source and supply-chain map to study — rather than a benchmark-comparable return stream or a track any reader can replicate net of costs. This is a critique of the record and the marketing frame, not a view on subscribing.
Not an AUM fee — access is a research subscription: Pro Deep Dives ~$474 first-year (promo); Advanced Market Signals ~$749 promo / ~$849 regular per year (io-fund.com, 2026). You pay for research + trade alerts, not fees on capital.