Goldman Sachs Bets on AI Infrastructure over 5% US Treasuries
Goldman Sachs executives argue that AI computing assets offer superior upside compared to government bonds yielding 5% in the current high-rate environment.
Key points
- Goldman Sachs executive Anshul Sehgal favors AI infrastructure assets over 5% US Treasuries due to higher growth potential.
- Sehgal argues that higher interest rates have increased savings, which are now funding AI data center and compute investments.
- The executive dismisses US debt sustainability fears, noting that interest payments accrue to capital rather than broad economic labor.
Government bonds currently offering yields near 5% have regained appeal for many investors, but a senior executive at Goldman Sachs believes the higher risk of AI infrastructure still presents a better opportunity. Anshul Sehgal, the bank’s global co-head of fixed income, currencies, and commodities, argues that the potential for multiplicative growth in computing power outweighs the modest upside available in long-term Treasuries.
Sehgal made these remarks on the bank's podcast, noting that while the Federal Reserve’s recent rate hikes have pushed long-term yields to their highest levels in over two decades, this has not dampened the momentum in the AI sector. Instead, he suggests that higher interest payments have provided savers with more capital, which is now flowing directly into AI infrastructure investments. Business Insider reported on this stance, highlighting a strategic preference for 'long compute' positions over traditional safe-haven assets.
Higher rates fuel AI spending
The prevailing view is that rising interest rates typically hurt growth-oriented sectors by increasing borrowing costs. However, Sehgal contends that for AI infrastructure, the dynamic is different. He points out that the increased interest income received by savers and institutions is being redeployed into the physical assets required for AI, such as data centers and neocloud providers. This creates a self-reinforcing cycle where monetary policy tightness indirectly supports the very sector that thrives on capital intensity.
This perspective distinguishes AI infrastructure from the broader equity market. Sehgal warns that while specific AI-related stocks may continue to outperform, the general stock market could still face headwinds from tighter monetary conditions. He emphasizes that the benefits of the current rate environment are not evenly distributed across all asset classes, making selective positioning in computing hardware and services a more precise strategy than a broad equity bet.
Debt concerns lack economic basis
Beyond asset allocation, Sehgal addressed widespread anxiety regarding United States fiscal deficits. He argues that investors have misinterpreted the nature of government spending, particularly the portion allocated to interest payments. According to his analysis, these costs do not stimulate broad economic activity or labor markets in the same way that new public spending does. Instead, they represent a transfer of income to capital owners, a distinction he believes is often overlooked in debt sustainability debates.
To support his argument, he noted that annual deficits have remained stable at 6% to 7% of GDP, while nominal GDP has grown at a similar pace over the last four years. This parallel growth, he says, limits the actual deterioration of the debt-to-GDP ratio. By contextualizing the fiscal data against economic output, Sehgal suggests that the long-term sustainability of US debt is less precarious than many market participants fear, further reducing the appeal of holding large positions in Treasuries as a hedge.
Trade-offs in asymmetric bets
Despite the bullish outlook, the strategy carries inherent risks. Focusing on AI infrastructure means taking on technological and valuation risks that are absent in government bonds. If the pace of AI adoption slows or if regulatory hurdles emerge, the 'multiplicative' upside Sehgal describes could stall. In contrast, while 5% Treasuries offer lower growth potential, they provide a fixed return with minimal credit risk, offering a safer floor for portfolios that cannot tolerate significant volatility.
The catch for investors following this advice is the need for precise timing and sector identification. The distinction between true AI infrastructure beneficiaries and broader tech stocks is becoming increasingly blurred, yet the financial outcomes may diverge significantly. While the executive is confident in the current setup, the market may not distinguish between these categories as cleanly, potentially leading to unexpected drawdowns in positions that were intended to be insulated from general equity market weakness.
