At first glance, last week seemed calm if you just looked at the main indexes.

But that was not really the case.

The S&P 500 gained 0.4%, and the Nasdaq Composite added just 0.1%. The Dow fell 0.6%. Meanwhile, the Russell 2000 gained 1.1%, outperforming all three large-cap indices.

This difference is more important than the headline numbers suggest.

During most of the AI rally, investors could follow the market by focusing on a few big tech companies. Last week showed that this approach is becoming less helpful. Money is now flowing into the infrastructure needed for AI, and rising oil prices have created another market-leading area.

The overall market hardly changed.

However, a lot was happening beneath the surface.

The S&P 500 and Nasdaq completed a third consecutive weekly advance, helped by softer inflation data earlier in the week. July CPI rose 0.1% month-on-month and 3.4% from a year earlier, down from 3.5% annual inflation in June. Core CPI increased 2.5% year-on-year.

Friday made things more complicated. U.S. retail and food-service sales dropped 0.6% in July, but they were still 5.0% higher than a year ago. While lower inflation and weaker spending eased worries about rate hikes, they also raised a tougher question: at what point does economic weakness become a real problem?

At the same time, Brent crude gained roughly 5.9% over the week amid continuing uncertainty around the Strait of Hormuz.

As a result, markets faced three competing factors: lower inflation, weaker consumer data, and higher energy prices.

Small caps

The Russell 2000 rose 1.1%, while the S&P 500 gained 0.4% and the Nasdaq just 0.1%. This does not mean the whole market has changed, but it does show that more types of companies are participating, instead of just the biggest tech firms leading again.

Energy

Oil prices made one of the biggest moves last week. WTI rose about 5.4% and Brent approximately 5.9% as uncertainty around Hormuz kept supply risks high.

Memory and storage

South Korea's July data showed memory-chip exports jumped 276.9% compared to the previous year. This sharp increase is another sign that AI demand is now as much about memory bandwidth as it is about computing power.

Technology momentum

Technology stocks lost ground by Friday, with semiconductors giving back some recent gains. The S&P 500 information-technology sector dropped 0.42% that day.

The consumer narrative

Retail sales in July fell 0.6% from the previous month. This does not mean consumers are in trouble, especially since sales are still 5.0% higher than in July 2025. However, it challenges the idea that the economy is picking up speed across every sector.

Anyone expecting a conventional semiconductor cycle

The industry's numbers are still far from normal. WSTS now expects global semiconductor sales to grow about 90% in 2026, reaching $1.51 trillion, with memory leading the way at around 250% growth.

Don't wait. Sign up for The Hustle to unlock our side hustle database. Unlike generic "start a blog" advice, we've curated 100 actual business ideas with real earning potential, startup costs, and time requirements. Join 1.5M professionals getting smarter about business daily and launch your next money-making venture.

Get the guide

The first stage of the AI boom was simple: more AI meant more demand for GPUs.

The next stage is more complex.

More GPUs mean more HBM is needed. More HBM calls for better packaging. Bigger clusters need networking, optical connections, storage, cooling, and power. All of this depends on having enough fabrication capacity and semiconductor equipment.

TSMC's July revenue reached NT$467.58 billion, up 44.7% from July 2025. Revenue for the first seven months of 2026 increased by 37.0%.

This is a key change. AI is no longer just a single product category. It is becoming a full industrial supply chain.

WSTS expects memory semiconductor revenue to increase roughly 250% in 2026 and exceed $800 billion. Memory alone is therefore expected to account for more than half of the organization's $1.51 trillion global semiconductor forecast.

South Korea's July export data reinforce that picture, with memory-chip exports up 276.9% year-on-year.

This does not mean memory has stopped being cyclical.

It just means this cycle is especially strong.

AI demand is now running into limits that software cannot quickly solve, such as wafer capacity, HBM supply, packaging, energy, and data center connections.

This is why investors are now looking beyond Nvidia and the biggest cloud companies and paying more attention to businesses involved in fabrication, lithography, memory, equipment, networking, and storage.

Software might be virtual,

but the bottlenecks are very real.

Technology stocks weakened before the CPI release, then rallied when July inflation came in at 0.1% month-on-month and 3.4% year-on-year.

So, a few weak days for tech stocks were more about investors adjusting ahead of a major economic report than about any real drop in demand for AI infrastructure.

Applied Materials dropped 5.1% on Friday even though its results beat expectations. This shows how high expectations can be a problem. Sometimes, even good results lead to weak stock performance if the market is already expecting something great.

This difference is important.

A drop in a semiconductor stock does not mean the whole industry is in trouble.

The opposite mistake is also important.

WSTS's numbers are impressive, but rapid growth makes future comparisons tougher. If the semiconductor market grows by 90% in one year, it will be much harder to sustain that growth from the higher starting point.

The cycle is now bigger.

But it is still a cycle.

The fact that the Russell 2000 gained 1.1% while the Nasdaq only rose 0.1% is an important detail.

The biggest tech companies did not have to fall for the market to broaden. Smaller companies just needed to do better.

This is a healthier way for the market to grow than relying only on the biggest companies.

The focus in semiconductors is shifting away from guessing which company will launch the best AI model. Now, it is more about figuring out where companies are investing their money.

WSTS now forecasts a $1.51 trillion semiconductor market in 2026 and approximately $1.9 trillion in 2027. Memory is expected to lead the current surge, while logic remains another major contributor.

This leads to a wider way of analyzing the market:

Compute benefits companies selling accelerators and CPUs.

Memory benefits HBM, DRAM, and NAND suppliers.

Manufacturing benefits advanced foundries.

Equipment benefits businesses that need to build additional capacity.

Networking and connectivity benefit from larger and more complex AI clusters.

There are now more opportunities.

But there are also more ways things could go wrong.

This week, the market’s focus shifts from inflation to housing, Federal Reserve updates, and consumer health.

Retail earnings are also getting more attention after July’s retail sales dropped unexpectedly.

The S&P 500 starts the week near record highs after three weeks of gains, but the broader economic picture is growing more complicated.