Deutsche Bank’s FX strategy team is out with a special report today that connects three things most people are still thinking about separately: the AI capital boom, asset tokenization, and the future of the dollar. The thesis from strategist Mallika Sachdeva is simple and uncomfortable. The US is going all-in on private foreign capital to fund both its AI buildout and its deficits, and blockchain rails are the tool being used to widen the funnel.
The scale of the US reach for capital this year has no precedent.
- AI venture funding has topped $400bn in 2026, running at three times last year’s pace, with over 90% of the biggest deals in the US
- The two largest US AI labs have raised a combined $217bn this year, with each valued near $1 trillion
- Hyperscalers (Google, Meta, Oracle, Amazon) averaged about $25bn a year in IG debt issuance over 2020-24. In 2026 YTD they’ve done roughly ten times that. Microsoft still hasn’t tapped debt markets
- Google did its first primary equity raise since the 2004 IPO in June, pulling in $85bn
- SpaceX listed near $2tn in June, the largest debut valuation ever, and reportedly earmarked 30% of the deal for retail
- US AI capex is expected around $800bn this year. China’s is closer to $100bn
The most important chart in the piece is on foreign equity inflows. Q2 2026 alone saw over $400bn of foreign equity capital come into the US, well above any prior quarter and well above what index performance alone would explain. Equity flows have now overtaken long-term debt flows as the main source of US capital account financing. That’s a critical change in the FX world and argues for more volatility as capital sloshes around global equity markets.
Deutsche calls it a “geopolitics-to-technology funding swap”: official, long-term, geopolitically-motivated buyers of Treasuries are fading, and private, shorter-term, tech-motivated equity buyers are taking their place.
Many traders I know have tuned out blockchain headlines but there are interesting things happening in stablecoins. Deutsche also argues that tokenization of US securities has moved from the fringe to the core of the plumbing.
The timeline:
- SEC’s Project Crypto launched July 2025, and folded into a joint SEC-CFTC framework in January 2026
- December 2025: SEC gave DTC a no-action letter, effectively blessing the idea that the same security can trade on traditional and on-chain rails with identical rights
- July 2026: DTC ran its first live tokenization tests with 40 firms. SPY was tokenized. JPMorgan posted tokenized collateral to meet a CME margin call
- DTC plans to launch its tokenization service by October, starting with the Russell 1000, ETFs and Treasuries
- NYSE is building a 24/7 platform with Securitize. Nasdaq is targeting full tokenization capability by 2027
Only about $40bn of real-world assets are tokenized today against a US asset universe north of $100tn. Estimates for the 2030s run from $2tn to $30tn, and Deutsche thinks those are low.
The Korea example is the one to file away. Korea is 2% of global GDP but accounted for 10% of the $740bn in foreign equity inflows to the US last year, largely retail, and largely because Korean brokerages lowered the access hurdles early with fractional trading and daytime access to US stocks. Deutsche’s point: if 24/7, instant-settlement, $1-minimum access to US assets goes global, that’s a powerful force.
What does it mean for the dollar?
Tokenization makes US assets more attractive (collateral velocity, instant repo, cheaper cost of capital) and more accessible. Both pull capital in. The problem is that the same rails that lower the hurdle for inbound capital lower it for outbound capital too. Deutsche’s conclusion is that the dollar is becoming riskier, more equity-sensitive, and directly leveraged to the AI race. The importance of equity flows is something I’ve increasingly observed since covid and now we’re seeing that growing into a mainstream view.
The report includes a scatter of currency-equity correlations against net international investment positions and notes that the safe-haven properties of the US dollar are eroding.
The tail risk is spelled out plainly: if AI proves uneconomical, or if the US starts to fall behind, the dollar is very exposed. Said differently, the dollar is increasingly a bet on US tech companies winning in AI.
China is playing a different game
The final section reframes the whole thing as a contest between economic models rather than AI models.
The US is running what Deutsche calls “open markets, closed tech”: throw the doors open to foreign capital, keep model weights proprietary, protect pricing power, reward shareholders. China is running “open tech, closed markets”: push out low-cost open-weight models, keep the capital account managed, fund capex from a trade surplus that is actually growing on AI-related exports.
China doesn’t need foreign capital, so it doesn’t need tokenization. In February the PBOC and CSRC tightened rules on RWA tokenization and offshore RMB stablecoins, while steering toward controlled e-CNY. Chinese labs have raised a fraction of what US labs have. DeepSeek’s $7.4bn was the largest round.
Chinese models are already more than half of US company token usage on OpenRouter. If the open-weight approach commoditizes AI and caps the excess profits of US labs, Deutsche argues it threatens more than valuations. It threatens the funding model that now underpins both the AI buildout and the US deficit, and the premise that open capital markets always deliver the winning innovation.
Strip out the blockchain language and this is a report about financing. The US has swapped official Treasury demand for private equity demand, is building infrastructure to make that swap bigger and faster, and has tied the dollar’s fortunes to whether AI earns a return. Europe is on board with tokenization but lacks the AI footprint. China has opted out because it doesn’t need the money.
For FX, the practical read is that USD is increasingly a bet on the Nasdaq, and the old assumption that the dollar rallies when equities fall deserves a hard second look.
