top of page

The AI trade unwinds (for now)

Nov 9
7 min read

November 2026



Dear Investors,


October saw a volatile rotation. Money left the AI trade and went into energy, staples, utilities and health care; the equal-weighted S&P 500 beat the cap-weighted index by the widest margin since 2009 while Nasdaq had its worst month since March 2025. Third-quarter earnings really kicked off the wave of volatility: Meta, Alphabet and Oracle all raised capex again to fund the AI buildout, and the market punished the depletion of free cash flows. Alphabet had its worst day on record, Oracle raised $20 billion at a discount and Intel's guidance miss turned a tech sell-off into a chip rout. Yields fell back below 5%, oil rose on Gulf supply risk, gold fell to a 12-month low, and bitcoin lost about 7% with the Binance hack adding forced selling at month end. The first week of November then compressed the month into two sessions: a post-election relief rally on 4 November and, on 5 November, the worst day for US stocks since April 2025, with the Nasdaq in correction and semiconductors down ~12% in a day.



Portfolio Activity


Our actively managed Digital Technologies Fund (DTF) was not insulated from these market events and saw most of the long-only positions detract. The drawdown (see our official factsheets) however was within expectations for the strategy, whilst still outperforming various other parts of the market including pureplay semiconductors and blockchain sectors. Our most notable portfolio activity is summarized below:


  • We rotated out of AI-infrastructure beta early. In the first half of October, we significantly reduced the miners and data-center sleeves, exiting Cipher Mining, Hut 8 and Galaxy Digital. The names we sold fell a further 12% to 27% by month end. We kept the operators however that own energized sites with capacity still to lease (Chart 1).

  • The book's composition shifted. Spot bitcoin is a smaller share of the fund than a month ago, with part of that exposure moved into calls. Hyperliquid, Block and the quantum sleeve are larger shares. Exchanges, brokers and tokenization are unchanged as a group, with adds to eToro and Figure on the mid-month low.

  • We raised cash tactically. Sales went into a cash buffer that we intend to redeploy as names reprice. This is a timing decision, not a change of view.

  • Our tail hedges paid, but they were sized relatively small. Puts on Nvidia, TeraWulf, Oracle and Cerebras were closed near the mid-month low of the underlyings. The covered calls on bitcoin and ether were bought back for a fraction of the premium received. We then bought puts on CoreWeave, the neocloud we believe carries the most refinancing risk relative to its valuation.

  • We moved part of our long exposure into calls. After Alphabet's 12.6% drop we bought long-dated calls on Alphabet and Amazon; after Intel's 29% drop, we bought long-dated calls on Intel and IBIT (Bitcoin ETF). With the Meta and AMD calls we already held, this keeps upside in the names we want to own while capping the downside if the correction runs further. We expect volatility in the near term but are constructive over the medium term, and calls let us hold both views at once.

  • One new position: GameStop. The company is consolidating hard, closing physical stores and moving into digital commerce and digital collectibles. That has hurt the top line by design, but it rebuilds the business on a leaner, better-capitalised base, and the valuation does not reflect the shift. It is no longer a headline name, which is exactly why we like it: we prefer companies that are out of the news but going through a real change in fundamentals, run by a CEO who is committed to keeping costs low.


The neocloud and chip sell-off. The group that led the market in September led it down in October: CoreWeave, Nebius, Cerebras and the miners-turned-data-centre operators fell 30% to 55%. The market sorted them by balance sheet; CoreWeave's bonds traded at a 15% yield and Gartner warned that most independent neoclouds will not survive on their own. On chart 1 we map the group on valuation per connected megawatt and unleased capacity, with bubble size showing refinancing risk. Our holdings sit in the cheap-with-upside quadrant with (relatively) little refinancing risk, and our current one short is the expensive, fully pre-sold name (CRWV). The sleeve is small, only about 8% ofthe fund, and we would rather be selective and tactical within this space.


Chart 1: Neocloud and miner universe, value versus repricing optionality, with the fund's holdings-weighted position. Source: BIT AM analysis, company filings, fund holdings as of 6 November 2026.
Chart 1: Neocloud and miner universe, value versus repricing optionality, with the fund's holdings-weighted position. Source: BIT AM analysis, company filings, fund holdings as of 6 November 2026.

Valuations as a buffer. Our equity book trades at ~20 times forward earnings versus ~23 for the Nasdaq-100 and ~36 for the Russell 2000. Our names have already been repriced while the indices at top-level may not fully reflect the sentiment shift yet. We acknowledge that more attractive valuations are no protection in a liquidation, but they can be a relative buffer once the market differentiates again.



Market View & Outlook


  1. Power is the bottleneck, which favours the owners of energized sites. Chip shipments and capex imply about 35 GW of new US data-centre capacity in 2027 and close to 60 GW in 2028; tracked construction, grid connections and turbine lead times suggest 24 GW and 30 GW are realistic. The gap opens next year and becomes a problem in 2028, when most planned capacity has not broken ground and the election puts another 37 GW of 2028 projects at risk of local pause. When compute is constrained by buildout delays, the scarce asset is a site with an interconnect, that is unleased and has repricing power. That is why we own the operators with power and capacity to lease, and why we are short the one that has sold everything forward.


Chart 2: US data-centre capacity added per year, chip-implied requirement versus realistic additions. Sources: Goldman Sachs, Jefferies and Aterio, Morgan Stanley, Wells Fargo, SemiAnalysis, SynMax, Cleanview.
Chart 2: US data-centre capacity added per year, chip-implied requirement versus realistic additions. Sources: Goldman Sachs, Jefferies and Aterio, Morgan Stanley, Wells Fargo, SemiAnalysis, SynMax, Cleanview.
  1. The correction was positioning, not fundamentals. AMD beat on every line on 3 November and fell ~15% two days later. Alphabet grew revenue 25% and lost half a trillion dollars of value in a day. The selling came from the holders as tech long/short hedge funds reportedly got battered on 5 November. The AI trade had become the industry's most crowded position, with the overlap among the largest multi-strategy funds high and rising, and when crowded positions lose, platforms cut leverage and sell regardless of price. Walleye had its worst month since March 2020 and is shrinking its equity business, Engineers Gate dismissed most of a stock team, Samlyn sold every stock outside financials, and Millennium pulled more than $1.5 billion from two external managers. The VIX jumped ~65% on 5 November, its biggest move in over two years.


    Our reading: a de-grossing of a crowded trade into a market with no natural buyer, amplified by angst over AI financing, possible data-centre regulation under a new Congress and a Fed that is not cutting. None of that changes where the compute is going but it changes the price at which we can own it. Forced sellers do not finish on a schedule, and we would deploy into this weakness rather than hide from it.


  2. Switzerland has its first Bitcoin ETF. On 20 October FINMA approved the first Swiss fund that holds bitcoin directly and is open to retail investors. Shortly after the news broke, we were excited to share that we are behind this ETF. The approval was covered by many news outlets including Finews, The Block, Cointelegraph, CVJ, SRP and many others. It closes a gap we wrote about in "The Case for New Alternatives": bitcoin has earned a place in the diversifier bucket of a portfolio, and until now there was no Swiss fund vehicle to hold it in.


  3. Crypto has fallen out of favour, and that is usually when it pays to build exposure. Nothing changed fundamentally in October: the first Swiss Bitcoin ETF was approved, the CFTC and SEC moved forward on market-structure rules, and institutional access keeps widening. What changed is attention. Capital and headlines went to AI, and crypto was left to drift through a month of ETF outflows and one exchange hack. We think periods of complacency and neglect are historically good entry points, and the return profile explains why. Chart 3 compares the skew of daily returns across asset classes. Our crypto portfolio is the only one with positive skew: its return distribution leans towards large up-days. The Nasdaq-100, the S&P 500 and gold all have the opposite profile, where the extreme days are mostly down-days, and long Treasuries sit near zero. For a portfolio that is what matters in a recovery: when the asset class catches a bid it tends to gap higher, and a position built during the quiet period captures that. This is why we keep a core allocation through the drawdown, add tactically on weakness, and hold calls rather than waiting for the all-clear.


Chart 3: Skewness of daily returns, the fund's crypto portfolio versus equity, gold and Treasury ETFs. Source: Bloomberg, BIT AM analysis.
Chart 3: Skewness of daily returns, the fund's crypto portfolio versus equity, gold and Treasury ETFs. Source: Bloomberg, BIT AM analysis.

For questions regarding our market views or our funds please contact us at info@bit-am.com.




Copyright © 2026 BIT Asset Management AG – All rights reserved.


This is an advertising document. This material has been prepared by BIT Asset Management AG for informational purposes only for the sole use of the intended recipient. It does not seek to make any recommendation to buy or sell any particular security or to adopt any specific investment strategy. This document does not contain information material to an investor’s decision to invest in a product. The information should not be regarded by recipients as a substitute for using their own judgment. Neither BIT Asset Management AG nor any of its affiliates, or their directors, officers, or employees, accepts any liability for any loss arising from the use of the information in this document. Data therein should not be relied upon as such information is subject to change, without notice, at the discretion of BIT Asset Management AG at any time. Investors in crypto assets are subject to the risk of total loss of the amount invested. Crypto assets are highly volatile and may fluctuate extremely in a short period of time. Crypto assets may become illiquid depending on trading platforms or investment product. Therefore, crypto assets are high-risk investments and you should not invest in this asset class unless you understand and can bear the risks involved with such investments. Although certain information has been obtained from sources believed to be reliable, we do not guarantee its accuracy, completeness or fairness. We have relied upon and assumed without independent verification, the accuracy and completeness of all information available from public sources.


BIT Asset Management AG is a manager of collective assets authorised by the Swiss Financial Market Supervisory Authority (“FINMA”) under the Financial Institutions Act (“FinIA").



 
 
bottom of page