Gavekal's Louis-Vincent Gave: China's Forced Pivot to Open-Source AI Is Building a "Digital Silk Road" That Could Commoditize OpenAI, Nvidia and the Entire AI Trade
Speaking on the Risk Reversal podcast with Dan Nathan and Peter Boockvar, the Gavekal Research CEO argues Washington's chip export bans backfired by pushing Beijing into an open-source strategy the US cannot easily counter, while oil, bonds, gold and the dollar all point to a regime change most investors have not priced in.
Louis-Vincent Gave, CEO of Hong Kong-based Gavekal Research, used a wide-ranging conversation on the Risk Reversal podcast to lay out a thesis that cuts against consensus on nearly every major macro trade of 2025: that China's AI models are advancing far faster than US investors appreciate, that oil and refining markets have entered a durable new range, that the bond bear market is structural rather than cyclical, and that the dollar's recent stability is masking a coming break lower. The interview, recorded as Gave was in New York meeting clients, ranged from the mechanics of Chinese naval strategy to why Marathon Petroleum's buyback math looks like "happy days" for energy investors.
China's Accidental Open-Source Advantage
The most striking argument concerns artificial intelligence. Gave contends that Washington's decision to cut off advanced chip sales to China did not simply slow Chinese AI development, it forced a strategic pivot that may now be working against American interests. "Historically, most Chinese companies are happy to throw money and human resources at things," Gave said. "This time, the US told them, 'You can't do it because we're not going to sell you the chips.' So they had to find workarounds. The only workaround was to open source it and have lots of software developers based everywhere around the world help you improve the product, which is completely anathema to the usual Chinese way of doing."
He frames the resulting competitive landscape through a memorable architectural metaphor: "OpenAI and Anthropic are essentially looking to build medieval fortresses with huge walls and say come into our fortress and you'll be safe here, to the Goldman Sachs of this world and the JP Morgans, etc. And China's building Dubai. It's not a fortress. It's everybody can come in, everybody can come out and you do whatever you want." The irony, he notes, is that America built the closed system while China, historically the closed society, was forced into openness purely because it lacked the compute to compete on Washington's terms.
The commercial implications are already visible. Gave cites Marc Andreessen's comment that three-quarters of the startups his firm sees are running Chinese large language models because they are essentially free, a dynamic he says extends "not just Singapore, everywhere in the world." His pricing analogy is blunt: Chinese labs are "selling you a Ferrari, but they're selling it to you for $10,000," and the open question is whether users tolerate the switch once the price reverts toward $300,000. He points to the recent shift from "token maxing to token minimizing" among US enterprises as early evidence that customers may not play along.
Semiconductors and Models Priced for a Monopoly That Won't Last
Gave's core warning to AI and semiconductor investors is about mispriced permanence. "Right now both of these are priced to essentially never be commoditized," he said, referring to both frontier models and leading-edge chips, "when we know that when China enters a space, it gets commoditized." He name-checked CXMT and Yangtze Memory Technology as Chinese memory producers investors in Micron should be studying closely, warning that Micron's roughly 85% gross margin business is squarely in Beijing's crosshairs. He also noted that semiconductors now make up roughly a fifth of global equity indices, up from 2% historically, forcing Asian fund managers who cannot legally hold more than 10% in any one name into "crappier semiconductor companies that I don't really want to own" simply to stay near benchmark, a dynamic he says will unwind badly once sales growth in the group rolls over.
Oil's New $65-to-$100 Range and the Refiners' Buyback Machine
On energy, Gave pushed back on the surprise many investors felt when oil retreated to the $60s despite the Iran conflict. He attributes the muted price action to China's enormous, underappreciated storage buffer, "probably closer to 1.8 billion" barrels versus an official 1.3 billion, plus expanded, partly discreet imports from Russia. The result, he argues, is a de facto price ceiling and floor set by Chinese buying behavior: "At 65, they buy as much as they can. 65 and below and at 100 they stop buying. So you end up, if that's the case, in a $65 to $100 range. I actually think that's great news for most energy companies."
He is more cautious on the disconnect between crude and refined products. Crack spreads have "blown out" to roughly $60 in the US, driven by Russian and Ukrainian strikes on each other's refineries, the loss of six major Gulf refineries including the second-largest in Bahrain during the Iran conflict, and China's decision to hoard refined products rather than export them. With Russia now halting product exports through July, a country that supplies roughly 11% of the world's diesel, Gave sees the refining bottleneck persisting. His preferred trade is the refiners themselves: companies like Marathon Petroleum and Valero, trading at all-time highs, generating what Boockvar called "enormous" free cash flow with no incentive to build new capacity. "Marathon is going to be buying five or ten percent of their stock for the next few years, every year," Gave said.
Empire-Building, Open Belts and the End of US Naval Control
Gave situates China's military buildup and Belt and Road initiative within a historical framework of empire, arguing Beijing has no territorial ambitions beyond Taiwan and is instead pursuing presence and trade infrastructure. The more consequential lesson from the Iran conflict, in his view, is that the US Navy no longer controls global sea lanes because "we now live in a world where you cannot protect billion-dollar ships with million-dollar missiles against $10,000 drones. The maths no longer add up." He calls this "a massive, massive shift" after 80 years of assumed American maritime protection, and argues it explains why China is investing heavily in its own navy, not to project offensive power but to protect its own shipping lanes now that it can no longer assume Washington either can or will do it for them.
OpenAI's Delayed IPO and the "Other Income" Problem in Tech Earnings
On the capital intensity of the AI buildout, Gave flagged OpenAI's decision to push its IPO to 2027 as a meaningful and underappreciated signal, not necessarily a positive one. "I would argue that OpenAI, from a strategic standpoint and the extent of its reach throughout this whole AI ecosystem, is too big to fail," he said, meaning its failure would cascade through the entire capex chain rather than trigger a government bailout. He also cited McKinsey estimates of $6.5 trillion in AI-related spending between now and 2030 and warned that markets are underestimating the maintenance capex embedded in hyperscaler commitments, including multi-decade data-center leases that function economically like debt.
Separately, Gave pointed to a distortion in reported earnings growth tied to hyperscalers marking up their stakes in AI labs. Citing research from macro strategist Kevin Muir, he noted that "other income," largely unrealized gains from investments like stakes in Anthropic, has been the fastest-growing line item in recent earnings, inflating headline numbers. "If you look at Q2 earnings that were up like 28%, if you take out the other income it was like 17%. If you take out semiconductors, it was like mid-single digits."
A Structural Bond Bear Market, Not a Cyclical One
Gave was unambiguous that developed-market bonds are in a multi-year structural bear market, tracing its origins to pandemic-era policy failures and now sustained by fiscal deficits he calls "ludicrous," including a 7% of GDP US deficit at full employment. He flagged Japan's 10-year JGB yield closing at a 29-year high and French OAT yields breaking out as evidence the pressure is global. The bigger risk, in his telling, is a potential unwind of Japanese capital: Japanese investors hold $3.5 trillion in foreign assets, roughly 80% of Japan's GDP, and Gave believes a JGB yield near 4.5% could eventually pull that capital home, "a damoclean sword over the head of the bond market everywhere." He argues the classic debt-to-GDP framework for judging sovereign risk is the wrong lens; what matters is the share of debt held by foreigners, which puts France and Britain, both around 30% foreign-owned, at greater risk than the US, whose debt benefits from reserve-currency status and "world-class companies" that keep foreign buyers interested regardless of valuation.
The Dollar's Coming Break and an Obsolete DXY
Gave believes the dollar's apparent stability, the DXY has been range-bound between 97 and 100, is deceptive because the index is two-thirds euro and roughly a quarter yen, "the index of a world of 25 years ago," with no meaningful China or commodity-currency weighting. He points to the renminbi appreciating roughly 50 basis points a month "like clockwork" against a backdrop of persistent yen weakness, calling the divergence a "rubber band" that will eventually snap in favor of Asian currency strength against the dollar. He also connects this to China's push to settle more than half its trade in its own currency and the slow build-out of its CIPS payments system, still in single-digit global market share but rising, arguing dollar de-basement "is not an event, it's a process."
Gold's Air Pocket and the Pivot to Commodity Stockpiling
On gold, Gave attributes the metal's recent pullback partly to a temporary shutdown of Dubai's physical gold market during the Iran conflict, a market he says represents roughly a quarter of global physical gold trade. More structurally, he distinguishes between the lessons of 2022, when Western seizure of Russian central bank and oligarch assets convinced central banks and wealthy individuals globally to diversify into gold, and the lesson of the Strait of Hormuz crisis, which he argues is different: countries need to stockpile actual commodities, not just gold or treasuries. He cited India's inability to secure fertilizer from China despite holding $700 billion in reserves as the case in point, and referenced fresh reports of the US Department of Defense stockpiling lithium and Germany building a strategic natural gas reserve as early evidence of a multi-year, inflationary trend in commodity demand. "I remain a bull on gold," he said, "but I'm much more bullish on the other commodities."
Portfolio Positioning: Financials and Cyclicals Over Crowded Tech
Assessing asset classes through fundamentals, momentum, positioning and valuation, Gave said tech's valuation and positioning look stretched with momentum possibly stalling, while financials, global banks including in Japan, Europe and China, screen attractively on all four measures. He linked the bank rally partly to steepening yield curves as central banks keep short rates pinned even as long-end yields rise, a dynamic he expects Japan to lead again just as it has with prior unconventional monetary policy experiments. Combined with his dollar-weakness view, Gave said his preferred equity exposure is overweight cyclicals and financials rather than the crowded AI and mega-cap tech trade.