Last month we argued the AI trade had absorbed 2026's attention and liquidity. Our view now goes one step further. The economics underneath the trade are cracking, and the same speculative capital that bid up AI equities will not sit still while the equity it lives inside reprices. It will look for the next asymmetric home. We think that home is crypto, and we think the rotation compounds with Bitcoin's own four-year cycle bottom in a window that runs the next 12 to 24 months.

The AI Trade's Assumptions Are Cracking

AI capital was underwritten on three assumptions: durable pricing power, hyperscaler margin expansion, and near-monopoly adoption of a small number of US frontier models. All three are already broken.

Start with scale. Microsoft, Alphabet, Amazon, and Meta guided to roughly $710 billion in combined 2026 capex, up from about $402 billion in 2025 — a 77% year-over-year step-up [1]. Amazon Q1 alone printed $44.2 billion, up 77% year over year, on a reiterated ~$200 billion full-year run rate. It is the largest coordinated private infrastructure buildout in history [2].

Pricing broke next. On April 23, 2026 OpenAI shipped GPT-5.5 at $5 input / $30 output per million tokens. One day later, DeepSeek released V4-Pro under an MIT license at $1.74 / $3.48 per million — roughly nine times cheaper on output. Claude Opus 4.7 sits at $5 / $25 on the same headline. V4-Flash prices as low as $0.14 to $0.28 per million tokens [3]. V4-Pro is cheap and it is capable. It scores 80.6 on SWE-bench Verified and leads LiveCodeBench at 93.5 — within striking distance of the closed frontier on the exact benchmarks enterprises use to make procurement decisions [3].

The market noticed. Chinese open-weight models went from under 2% of OpenRouter token consumption in late 2024 to more than 50% by June 2026, with US models collapsing from roughly 70% to roughly 30% over the same window. DeepSeek alone now runs about 16.3% of platform token volume — more than any single OpenAI, Anthropic, or Google offering [4].

Enterprise buyers stopped being loyal. Uber exhausted its entire 2026 AI budget in four months, largely on Claude Code, with heavy users running $500 to $2,000 per engineer per month [5]. Lindy migrated 100% of production traffic from Anthropic to DeepSeek V4 and told its investors the switch was business-transforming [6]. Enterprises are moving on price alone. Not in benchmarks — in production.

The accounting is the last leg. Hyperscalers depreciate GPUs over a 5-to-6-year useful life, but the economic life of frontier-generation chips is closer to three years. Evercore and Bank of America have both flagged the gap, with implied EPS inflation in the 10–15% range [7]. Amazon's ~$200 billion capex plan likely pushes 2026 free cash flow negative [7]. Capex-to-sales climbs to roughly 86% at Oracle, 54% at Meta, 47% at Microsoft, and 46% at Alphabet [8]. The top ten S&P 500 companies now represent about 40% of the index — well above the roughly 27% concentration at the dot-com peak. The margin assumption is already broken. The capex bill is not.

Bitcoin Has Already Decoupled From the AI/Nasdaq Complex

Bitcoin has already decoupled. The market got there ahead of us, and it went the wrong way first.

Through 2025 and into early 2026, Bitcoin traded as a high-beta Nasdaq. The 30-day rolling correlation between BTC and the Nasdaq 100 hit 0.96 in April 2026 — a level at which roughly 92% of Bitcoin's variance tracked tech equities [9]. By early June it collapsed toward zero. By early July it read −0.84 against Nasdaq and −0.82 against the S&P 500, the most negative in over two years [10].

The return gap opened in step. Bitcoin sits roughly 50% below its October 2025 peak of $126,198, while the Nasdaq 100 recovered more than 30% off its April lows and is closing on new highs [11]. Call it a 70-point return gap in eight months.

The decoupling has hurt Bitcoin, not helped it. Anyone selling the easy version of this memo — "AI weakens, so crypto rallies" — is asking you to believe something the last two quarters have already disproven.

The Bridge: From Bearish Decoupling to Bullish Rotation

If falling AI valuations alone drove crypto strength, we would see it already. We do not. Bitcoin is down. So the rotation thesis needs a real bridge, and it has one.

The decoupling changed the channel through which crypto prices — the direction of the pricing regime has not reversed, the input has. Crypto no longer prices off tech-sector sentiment. It prices off its own liquidity conditions. LSEG data through 2026 shows BTC's synchronicity with the 10-year US Treasury yield has strengthened even as the Nasdaq link has broken. The pricing wire has been rerouted from equity beta to macro liquidity beta [9].

The catalyst we care about is what an AI-driven de-risking event forces policy and dollar liquidity to do. Falling AI stocks are the trigger. The transmission runs through liquidity. When index-heavy AI names carry 40% of the S&P and their earnings assumptions get marked down, the Fed and Treasury do not watch that from the sidelines. They respond through the liquidity channel. That is the same channel Bitcoin now prices off.

Crypto is where the money lands because crypto is cheap and loaded. Bitcoin's MVRV Z-Score sat at 0.37 in mid-July 2026 — roughly a quarter of its long-run average near 1.7, and in the same zone that preceded every major recovery in 2015, 2018, and 2022 [12]. Stablecoin supply sits near $312 billion, down only about 3% from the May peak of $322 billion. Compare that with the 26% collapse in the 2022 stablecoin market [13][14]. The dry powder did not leave. It sat down.

Our Forwarded Liquidity Momentum framework is the third piece. Global liquidity stands at a record $192.4 trillion, but the Shadow Monetary Base — central-bank liquidity plus bond-market collateral — is contracting at a 6.3% three-month annualized rate [15]. The headline is being carried by narrow Fed operations and suppressed bond vol. Underneath, the system is shrinking. We flagged August–September 2026 as the pivot read. We are still on that clock.

The same regime that produced underperformance on the way down enables sharp repricing on the way up. When liquidity turns and BTC prices off liquidity, the tape does not need a rebuilt Nasdaq correlation to move. It needs the shadow base to inflect.

Then there is the cycle, which skeptics have called dead every cycle and been wrong every cycle. The bull phases of 2013, 2017, and 2021 lasted 1,050, 1,071, and 1,064 days. The bear phases that followed each ran almost exactly 52 weeks. Bitcoin's Q4 2025 top on October 6, 2025 at $126,198 fits the pattern to the day [16]. On June 4, 2026, Bitcoin tagged its 200-week moving average near $61,000 for the first time this cycle — a line that has coincided with every prior cycle low [17]. Long-term holders now control roughly 16.64 million BTC — about 78% of circulating supply, an all-time high [18].

The cycle's organic recovery and the exogenous AI-liquidity rotation arrive in the same window. That is what gives this cycle the room to exceed prior cycles in percentage terms. One catalyst you can time. One you cannot. Both point up.

One outside data point supports it. Bitwise's July 2026 staking work argues the biggest bullish signal in crypto right now has little to do with price and everything to do with fundamentals — staking demand at record highs, on-chain usage climbing while fees compress, and institutions quietly deploying capital for yield [19]. Blockchain Capital's Aleks Larsen and Spencer Bogart put a hard number on the same gap: in 2021 more than 70% of user fees on chain went to infrastructure. In 2025, application-layer fees surpassed infrastructure fees for the first time — value has already migrated up the stack, even as prices have not followed [33]. The divergence between price and fundamentals is now wide enough that multiple serious desks are calling it out.

Why Crypto Specifically — The Behavioral Case and the Numbers

Most of the capital bidding up AI equities was never buy-and-hold money. It was asymmetry-seeking, high-volatility, venture-style capital temporarily parked in a public-equity wrapper. That kind of capital does not rotate into investment-grade bonds or defensive equities when its host cools. It looks for the next place with the same profile. Crypto has no single-company earnings risk, trades 24/7 in globally accessible venues, and has three prior cycles of documented asymmetric outcomes on the tape.

The counterfactual is thin. Rob Arnott projects S&P 500 total returns of just 3.1% annually over the next decade — barely ahead of inflation [20]. Goldman's Kostin sits near the same 3% mark. Speculative capital does not accept beta returns. It leaves.

The consensus cluster tells us where the mainstream is anchored. Citi's base case for the next 12 months sits at $143,000, with a bull case of $189,000 and a bear case of $78,500 [21]. JPMorgan sits at $150,000 to $170,000. Standard Chartered is at $150,000. Bernstein holds $200,000 and calls the current setup "the weakest bitcoin bear case in history" [22]. Published forecasts across the tape range from $38,000 to $250,000 for 2026 outcomes — the dispersion is wide because most models are not pricing the same catalyst.

We think the $143K–$170K cluster understates the outcome because none of it explicitly prices the AI-capital rotation. Our base case sits above consensus at $170,000 to $200,000+. Bernstein's $200,000 is, in our view, closer to the true setup than the mid-cluster targets.

We are explicit about the contingency. If the AI-capex reset unfolds on the Section II timeline, our base case is $170,000 to $200,000+. If it does not, we default to the cycle-only case of $143,000 to $170,000 — the consensus range that cycle mechanics can deliver on their own. The floor is the same either way. We anchor Bitcoin's cycle low at $50,000 to $55,000, aligned with the aggregate realized price near $54,000 that has been the deeper structural support in every prior cycle [16].

The Structural Amplifier: AI Agents Will Run on Crypto Rails

There is a second reason capital lands here, and it runs on a longer clock than the cycle.Autonomous AI agents need to pay for things, and web infrastructure has never had a native machine-payment layer. In May 2025 Coinbase revived HTTP 402 — the dormant "Payment Required" status code — into a working protocol called x402. Servers respond to an agent's request with a 402 and a payment payload. The agent signs a stablecoin transfer and retries. Settlement is embedded in the round trip. No accounts. No API keys. No human approval.

By late April 2026 the protocol had processed 165 million transactions across roughly 69,000 active agents and about 480,000 agents transacting since launch. Cumulative volume sat near $50 million on an average call value around $0.31 [23][24]. On July 14, 2026 governance formally transferred to the Linux Foundation. Premier Members now include Visa, Mastercard, American Express, Stripe, Ripple, Google, AWS, Cloudflare, Circle, Adyen, Polygon, the Solana Foundation, Shopify, and Coinbase itself [25]. This is the current payments oligopoly signing on to a stablecoin-native standard.

The broader stack is filling in. AWS Bedrock AgentCore Payments integrates Coinbase and Stripe. Solana Foundation and Google Cloud are building Pay.sh. Circle's Agent Stack ships. Mastercard's AP4M protocol records agent permissions on Polygon, Solana, and Base. Every serious infrastructure vendor is placing its agentic-commerce bet on crypto rails.

The builder base matches the infrastructure. Headline crypto developer counts fell sharply through the bear, but the decline was concentrated in newcomers. Developers with two-plus years of experience hit an all-time high — up 27% year over year, now producing roughly 70% of all crypto code. That experienced core is concentrating on exactly the primitives agentic AI needs: payments, identity, permissioning, and settlement.

We are sober about dollar volumes. x402 processed only about $24 million in a rolling 30-day window as of mid-July 2026 [25]. This is a two-to-four-year structural tailwind on top of the cyclical thesis. The near-term trigger remains the AI-capex reset and the liquidity response it produces. Rails are what compound the bounce.

Supporting Structural Tailwinds

Four currents run underneath the thesis.

The GENIUS Act, signed in 2025, is in implementation and gives the US its first federal framework for payment stablecoins — 1:1 reserve backing, monthly audits, and dual federal/state oversight [26]. The CLARITY Act passed the House in 2025 and cleared Senate Banking 15-9 in May 2026. Treasury Secretary Scott Bessent is pushing for a full-Senate vote [27]. Nasdaq's own analysis argues passage would make it "much easier for Wall Street firms to introduce new crypto-related financial products" [28]. Regulatory clarity is an independent catalyst — one that can fire even if the AI reset stalls.

Institutional flows are already voting with capital. Tokenized real-world assets on public blockchains grew from about $6 billion in early 2025 to roughly $19.4–23.6 billion by March 2026 [29][30]. More than triple in fifteen months. RWA protocols overtook lending to become DeFi's largest sector while overall DeFi TVL was flat to down. Institutions chose infrastructure over beta. BlackRock and Kraken are deploying through on-chain vault standards. Robinhood built a Layer-2 purpose-built for tokenized RWAs [31]. Bitwise's staking work adds a parallel thread: institutions are quietly using staking as a yield layer on core crypto holdings — committed capital that does not exit on price weakness [19].

Corporate treasuries are the structural bid underneath all of it. As of July 2026, 175 listed companies collectively hold 1.28 million BTC [32]. That figure held through the worst month of ETF outflows since daily records began. It is a demand-side floor that operates independently of retail sentiment, ETF flows, and the rotation trade itself.

Stablecoins are the fourth current, and the one that turns a regulatory tailwind into a cash-flow mechanic. Blockchain Capital's analysis puts stablecoin velocity around 120x — the average stablecoin dollar turns over 120 times per year — and every $1 billion of net new issuance drives roughly $19 million of annual onchain protocol revenue downstream in lending venues, DEXs, and derivatives [33]. Supply sits near $312 billion today. Blockchain Capital projects a path to $2 trillion by 2030 [33]. On that math, the revenue accruing to onchain applications alone runs into the tens of billions annually by decade-end — before counting what accrues to Tether, Circle, or the exchanges. This is what happens when the GENIUS Act framework in the paragraph above meets a network already spinning at 120x velocity.

House View

Bitcoin's floor sits at $50,000 to $55,000, aligned with the aggregate realized price zone that has been the deeper structural support in every prior cycle. Base case is $170,000 to $200,000+ if the AI-capex reset unfolds on the Section II timeline. Cycle-only fallback is $143,000 to $170,000 — the consensus range that cycle mechanics deliver on their own. The window we care about is 12 to 24 months.

The reason this cycle can run larger than prior ones is the AI-agent rails convergence. That amplifier does not need to arrive in the next quarter. It only needs to remain on trajectory. If it does, the upside case is the base, not the ceiling.

We will revisit this view in future monthly memos as the confirmation window in the Forwarded Liquidity Momentum framework prints, as the CLARITY Act path resolves, and as the hyperscalers' Q3 and Q4 numbers force the market to reckon with the capex reset it has so far chosen to ignore.

Disclaimer

This material is provided for information purposes only and does not constitute investment advice, a recommendation, or an offer or solicitation to buy or sell any financial product or service in any jurisdiction.

The views expressed are those of the author as at the date of publication and are subject to change without notice. While reasonable care has been taken, no representation or warranty, express or implied, is made as to the accuracy, completeness, or reliability of the information contained herein, and no reliance should be placed upon it.

This material has not been reviewed or approved by the Dubai Financial Services Authority, the British Virgin Islands Financial Services Commission, or the Cayman Islands Monetary Authority, or any other regulatory authority.

This material is not intended for distribution to, or use by, any person in any jurisdiction where such distribution or use would be contrary to local law or regulation. In particular, this material does not constitute a public offer of securities in the British Virgin Islands or the Cayman Islands.

Any investment or service referred to herein will be available only to persons who meet the relevant eligibility criteria under applicable laws and regulations, including, where applicable, Professional Clients (as defined by the Dubai Financial Services Authority) or other equivalent categories of sophisticated or professional investors under applicable law.

Investments involve risk, including the possible loss of capital. Past performance is not a reliable indicator of future results.

“Axys” is a trading name used by a network of independent businesses. Axys does not operate as, and is not, a separate legal entity in any jurisdiction and has no distinct legal personality. Each business operating under the Axys name is independently owned and operated.

Axys Capital  Ltd is authorised and regulated by the Dubai Financial Services Authority. Axys Investment Management Ltd is authorised and regulated by the British Virgin Islands Financial Services Commission.

References

Analysis Atlas — https://analysis-atlas.com/research/hyperscaler-ai-capex-2026/

Value Add VC — https://valueaddvc.com/blog/big-tech-ai-capex-in-2025-microsoft-google-meta-amazon-and-the-spending-race

BenchLM — https://benchlm.ai/blog/posts/deepseek-v4-vs-claude-opus-4-7-vs-gpt-5-5

Signal — https://www.readsignal.io/article/chinese-open-weight-models-openrouter-token-dominance-2026

Fortune — https://fortune.com/2026/05/26/uber-coo-ai-spending-tokens-claude-code/

Forkable — https://www.forkable.io/p/ai-startup-ditches-anthropic-for

Evercore ISI research summary via Value Add VC — https://valueaddvc.com/blog/ai-hyperscaler-capex-compared-why-microsoft-google-meta-and-amazon-are-all-spending-at-once

CreditSights — https://know.creditsights.com/insights/tech-raising-hyperscaler-capex-2026-estimates/

Gate.com — https://www.gate.com/blog/has-btc-decoupled-from-the-nasdaq-2022-2026-rolling-correlation-data-and-the-asset-identity-inflection-point

BTCtiming — https://btctiming.com/en/blog/btc-nasdaq-correlation-flips-negative

Taurex — https://www.tradetaurex.com/news/coffee-charts-with-connor-bitcoin-is-getting-left-behind/

Coincub — https://coincub.com/blog/value-bitcoin-during-crash/

StablecoinInsider — https://stablecoininsider.org/stablecoin-liquidity-report-2026-supply-volume-venue-depth-and-defi-utilization/

CEX.IO — https://blog.cex.io/ecosystem/q2-2026-stablecoin-report-35673

Axys Research — https://www.axysinvestment.com/research/forward-liquidity-momentum

CryptoCompass — https://cryptocompass.com/articles/bitcoin-s-200-week-average-break-tests-the-four-year-cycle

Spoted Crypto — https://www.spotedcrypto.com/bitcoin-200wma-bear-bottom-signal-2026/

Bitcoin News — https://news.bitcoin.com/bitcoin-long-term-holder-supply-all-time-high-2026/

Milk Road / Bitwise — https://www.youtube.com/watch?v=3FPhmpdk2U0

Fortune — https://fortune.com/2026/03/17/spx-stock-market-future-returns-outlook-warning/

Yahoo Finance / 247 Wall Street — https://finance.yahoo.com/news/citigroup-143k-bitcoin-call-2026-173726131.html

CoinGecko roundup — https://www.coingecko.com/learn/bitcoin-price-predictions-expert-forecasts

x402 Foundation / Coinbase disclosures via Quicklizard — https://quicklizard.com/blog/the-settlement-layer-how-x402-completes-the-agentic-commerce-stack/

Datawallet — https://www.datawallet.com/crypto/x402-protocol-explained

Linkmerica — https://linkmerica.com/research/coinbase-agentkit/

Decentralized Finance — https://decentralized-finance.io/article/us-crypto-regulation-2026-roadmap/

CNBC — https://www.cnbc.com/2026/05/14/clarity-act-congress-crypto-senate.html

Nasdaq — https://www.nasdaq.com/articles/2-big-crypto-regulations-dropping-q2-2026

Earnpark — https://earnpark.com/en/posts/tokenized-real-world-assets-hit-19-8b-all-time-record-and-why-yield-seekers-should-pay-attention/

Blockster — https://blockster.com/rwas-in-january-2026-from-hype-to-a-core-defi-vertical

Lagoon — https://lagoon.finance/blog/state-of-onchain-vaults-2026

SatsIntel — https://www.satsintel.io/en/news/bitcoin-market-sentiment-july-2026-record-etf-outflows

Bankless / Blockchain Capital (Aleks Larsen and Spencer Bogart) — https://www.youtube.com/watch?v=lyIKcpa582I