Goldman Sachs: AI Could Ignite U.S. Inflation; Memory Price Surge Is the Key Driver

Introduction

In recent years, the rapid advance of artificial intelligence (AI) has been profoundly reshaping the global economic landscape. From large language models to autonomous driving systems, every step forward in AI depends on massive data-processing and storage capabilities. Yet just as the industry celebrates AI's productivity potential, Goldman Sachs has issued a sobering forecast: AI could become the trigger for a new round of U.S. inflation, and soaring memory device prices are the core driver of that risk. This article breaks down Goldman Sachs' view, explores the causal chain between AI and inflation, and assesses the possible impact on the macroeconomy and investment markets.

Main Text

1. AI's Hunger for Compute: Storage Demand Surges

AI model training and inference depend on huge amounts of data. Every parameter update and every model iteration relies on high-performance storage systems. In recent years, with the spread of generative AI, edge computing, and big-data analytics, demand for solid-state drives (SSD), high-bandwidth memory (HBM), and enterprise storage equipment has grown exponentially.

Industry reports show that global AI server shipments rose more than 30% year on year in 2023, and the storage capacity required by each AI server is often several times, or even dozens of times, that of a traditional server. Take NVIDIA's H100 GPU acceleration platform as an example: the HBM capacity paired with it has increased from 80GB in the previous generation to 144GB, and it will keep rising as more advanced chips are released. Orders at storage suppliers such as Samsung, SK Hynix, and Micron are already booked several quarters ahead, and supply shortages are now a reality.

2. Soaring Memory Prices: How the Shock Passes Through the Chain

A structural squeeze on supply and a sudden surge in demand have together driven memory prices sharply higher. Since the second half of 2023, NAND Flash and DRAM contract prices have risen for several consecutive quarters, with some specs jumping by more than 50%.

Goldman Sachs says in its latest research that this price spike is not just a normal cycle, but a structural shift with a longer-term trend. There are three reasons: first, memory-chip manufacturing is approaching physical limits, and the cost and difficulty of increasing 3D NAND layer counts are rising at the same time; second, after the 2022 price collapse, the major memory makers have generally adopted strict capex discipline, limiting new capacity; third, AI's special requirements - high bandwidth, low latency, and large capacity - demand far higher specifications than consumer products, pushing unit costs up further.

Those costs will eventually be passed from server makers and cloud providers down to enterprises and end consumers. When the cost of building AI infrastructure rises sharply, companies that rely on those compute services will inevitably raise the prices of their products or services, creating upward pressure on the overall price level.

3. The Inflation Transmission Path: From Hardware Costs to Service Prices

Goldman's model shows that the effect of memory prices on core inflation does not show up directly in the computers-and-peripherals component of the CPI. Instead, it spreads through broader channels. Specifically, the path includes:

1. Higher cloud prices: Amazon AWS, Microsoft Azure, and Google GCP are the biggest buyers of AI compute and also the biggest consumers of memory chips. As storage costs rise, cloud providers have little choice but to raise rental fees for computing power and storage. Those costs are passed on to enterprise customers and ultimately consumers, pushing up prices in e-commerce, streaming, and online advertising.

2. Cost pass-through in data-center expansion: The AI boom has pushed global data-center construction into overdrive, and storage equipment accounts for 30% to 40% of data-center IT equipment costs. With construction, power, and hardware costs all rising, operators must transfer the burden to customers through price-adjustment clauses in long-term contracts.

3. Delays in enterprise AI rollout and substitution effects: For small and mid-sized companies, high storage costs may delay AI deployment or force them to choose even more expensive in-house solutions, slowing the deflationary benefits that productivity gains could bring. In the long run, if AI fails to reduce operating costs on schedule and instead becomes a source of higher costs, it could create cost-push inflation.

4. A Historical Parallel: From Chip Shortage to Memory Shortage

Looking back at the global chip shortage from 2020 to 2022, automakers, consumer electronics makers, and other industries were forced to cut output because of chip shortages, which eventually pushed up the prices of new cars, appliances, and smartphones and became one of the major drivers of inflation in Europe and the U.S. Goldman Sachs now sees a similar pattern emerging in memory, except this time the driver is not pandemic demand but the structural demand created by AI.

Compared with that chip shortage, the memory price surge may last longer. The reason is that AI compute demand is nowhere near its peak, and memory technology upgrades have long replacement cycles, so there is little chance of a near-term substitute product that could ease prices. Geopolitical factors are also adding uncertainty to the supply chain. U.S. export controls on Chinese semiconductor technology are constraining some memory capacity plans and worsening the supply squeeze.

5. Macro Impact Assessment: A New Challenge for the Fed

If memory prices stay elevated, U.S. core inflation could rebound in the second half of 2024 through 2025. For the Federal Reserve, which is trying to achieve a soft landing, that is clearly a difficult variable. Goldman modeled two scenarios:

• Base case: memory prices peak in Q4 2024 and then gradually ease, with the impact on overall inflation around 0.1 to 0.2 percentage points, allowing the Fed to keep its current rate-cut path.
• Pessimistic case: if AI demand beats expectations and supply improves slowly, memory prices could rise another 20% to 30%, pushing core inflation up by 0.4 to 0.6 percentage points. In that case, the Fed may be forced to pause cuts or even consider raising rates again, which would hit stocks and bonds.

It is worth noting that this is a form of technology-driven inflation, different from traditional overheating demand or wage-led inflation. Monetary policy has limited power to control it, because the root problem is a supply-side bottleneck. If the Fed tightens too aggressively, it could suppress normal activity in other parts of the economy without directly solving the memory-chip shortage.

6. Investment Takeaways and Risk Management

For investors, Goldman Sachs' view points to two themes: first, upstream memory suppliers are gaining pricing power and their stocks may continue to benefit; second, downstream companies that depend on AI compute, such as software firms and cloud providers, may face margin pressure. Investors should also watch memory-industry expansion plans and technical breakthroughs, such as storage-class memory (SCM) and optical interconnect technology, which could help bring prices back down.

At the same time, defensive assets such as gold and inflation-linked bonds (TIPS) may attract flows in an uncertain environment. Portfolios should modestly reduce exposure to rate-sensitive growth stocks and shift toward more defensive allocations.

Conclusion

Goldman Sachs' argument that AI could ignite U.S. inflation offers an important lens for examining how technological progress interacts with the macroeconomy. Soaring memory prices, as the core link in that chain, are affecting far more than the semiconductor industry itself; through the construction and operation of AI infrastructure, they are spreading across the broader economy. In the short term, markets need to watch for price pressure from supply bottlenecks. In the long run, only faster innovation in memory technology and a rational expansion layout can truly resolve this structural contradiction. For policymakers and investors alike, understanding and managing this new inflation risk has become a task that cannot be ignored in future economic governance.