Oil prices above $100, massive investment in artificial intelligence and rising bond yields. The markets are facing several sources of tension simultaneously. For Silvano Marchesi, it is now the trajectory of real interest rates, US debt and the financing of the AI cycle that will determine what happens next.
Brent crude has jumped above $109 a barrel, feeding directly into inflation and rate concerns. How much of your portfolio positioning right now is a direct response to the oil shock rather than to underlying growth fundamentals?
We have no direct portfolio response to the oil shock at this stage. Our positioning was already built around these risks well before the current move in oil. For years, we have been tactically underweight nominal assets and overweight real assets, especially gold, based on low real rates first, then supply chain disruptions, growth, monetary policy and inflation.
We trimmed our gold position earlier this year after its strong performance, but the broader positioning remains unchanged. Today, we are still underweight bonds and duration.
The events in the Gulf have been a surprise, but the oil shock has reinforced rather than changed our positioning. Oil is currently pricing in both a supply shock and demand destruction. If demand were not weakening, I believe oil could already be at $140 or higher.
If yields, particularly real yields, were to rise significantly further and we saw a clear shift towards stronger demand destruction and deflationary risks, we would become more interested in increasing duration. At that point, the compensation offered by bonds would finally become attractive after a decade of very poor returns.
Dario Amodei's call for a slowdown in AI development has just had quite an impact on tech stocks. To what extent would you consider this as a genuine inflection point for the AI trade?
For me, the real test would come down to two numbers. First, the capex guidance from the hyperscalers. Second, how that capex is being financed.
There is another way of looking at Amodei's comments. Anthropic and OpenAI are among the companies carrying very high expectations and potentially very high IPO valuations. At the same time, the gap between the leading models is narrowing as open source models, Google and others continue to make progress.
If the current leaders start losing their edge, it becomes much harder to justify valuations that depend at least partly on their leading position. That is why I am not entirely convinced that safety is the only concern behind calls for a slowdown. Safety is obviously a genuine issue, and AI progress raises important questions. But asking everyone else to slow down also conveniently asks your competitors to slow down.
The real test is what these companies actually do. Do they reduce their own spending, hiring or release schedules? And do the hyperscalers revise their capex plans? The numbers for 2026 and 2027 still point to enormous investment and substantial negative cash flow for the next couple of years. Ultimately, those investments have to be monetised. We need to see aggregate hyperscaler free cash flow start growing again in order to stay cautiously optimistic on markets.
If AI safety concerns genuinely translate into slower capex growth among hyperscalers, what happens to the earnings assumptions that have justified current valuations for names like Nvidia?
What would slow this investment cycle, in my view, is not safety concerns but cost. And cost is rising from two directions: through interest rates, meaning financing – and through scarcity in everything a data centre needs beyond chips, from grid connections to transformers to cooling.
Nvidia is an interesting case because its actual earnings growth has now caught up with part of the valuation concerns we had two or three years ago. Earnings visibility today is much better – but it depends directly on hyperscaler capex. And as long as demand supports prices, more expensive components are passed through; if the scarcity eases, they will weigh on Nvidia's margin as well.
If capex genuinely grows more slowly – particularly because monetisation is taking too long or projects are being cancelled – there will inevitably be earnings revisions. And those would not affect Nvidia alone, but the entire supply chain and all the sectors that live off this investment cycle.
What worries me most in this context, however, is circular financing. If one company invests in another and that company then uses the money to buy the first company's products, the same dollar can effectively be counted twice – once as investment, once as revenue. If this form of financing is expanded further to keep the cycle alive, that would be a significant risk to its sustainability.
The 10-year Treasury yield just crossed 5 per cent for the first time since 2023, a level some analysts flag as a threshold where markets could tip into real stress. At what point do higher-for-longer government borrowing costs stop being a headwind and start being a systemic risk in your view?
I don't think it will become a systemic risk because the US Treasury and the Federal Reserve will intervene before reaching that point.
Comparing today's nominal yields with levels seen in 2007 or 2003 can be misleading. The amount of debt has increased dramatically. High yields with relatively low debt can be manageable. High debt combined with high yields is where the problem becomes much more serious.
US gross interest expense is now around $1.27 trillion a year, compared with roughly $450 billion in 2021, and it is still growing at around 10 per cent a year. Around one third of tradable debt matures within the next 12 months. Then you have 2027 and 2028, when roughly $6 trillion of maturing debt, issued at much lower rates, will have to be rolled over. Refinancing is therefore becoming increasingly expensive.
For me, the key issue is less a specific market threshold than a political pain threshold. We are likely to see increasingly large interventions. We have already seen Treasury buybacks of long-dated paper increase sharply, while yields continued to rise.
Ultimately, if the situation deteriorates far enough, yield curve control becomes the most probable policy response. The alternatives are higher taxes, austerity, or default, which are politically way more costly than just inflating the debt away, as it does not require parliamentary approval.
With oil, AI valuations and bond yields all moving against risk assets at once, which of these three worries you most right now?
The bond market.
The bond market is different because it changes the rules for every other asset class at the same time. It determines the discount rate for equities, the financing cost of AI capex, mortgage rates and the economics of other investment projects. Most importantly, it determines the real return on what is supposed to be the safest asset in the world, US Treasuries.
I am also concerned that the policy response itself could become the risk. There is already political pressure on the Fed to bring yields down, while fiscal measures could add further inflationary pressure. That raises questions about Fed credibility and the possibility of fiscal dominance, where financing government debt takes priority over controlling inflation.
Oil remains the driver of the price chain. With the Saudi East-West pipeline shut down, what is missing is physical barrels, not just a sense of security. As long as the Gulf conflict remains unresolved, we expect high and volatile prices that feed into higher producer and consumer prices and, through inflation, weaken demand. The same price that drives inflation ultimately destroys the demand that keeps it high. That is the tipping point we are watching closely.
Given this backdrop, what are the asset classes or regions you're actively rotating into or out of to prepare for the coming quarter?
Over the past few years, we have increasingly built alternatives into portfolios as a source of diversification. Cat bonds are a good example because their correlation with equities and bonds is very low. Managed futures are another. Our current focus is therefore less about making one large directional bet and more about diversifying away from the main risks.
We are also looking at Nordic bond markets, although we are not invested there yet. We are looking for markets that are less exposed to fiscal dominance and financial repression. In our view, the Nordic countries are generally more disciplined on these issues.
We are also looking to increase our gold overweight following the recent correction. If demand destruction becomes the dominant force and the liquidity pump is reopened, I would expect gold to benefit strongly.
So, in simple terms, our focus is turning to Nordic bonds and again on alternatives, especially gold.