The S&P 500 is heading for a weekly loss of more than 1.5% despite positive developments including oil falling roughly 10% as the Strait of Hormuz reopened, May core PCE coming in broadly as expected, and ten-year Treasury yields easing by more than 10 basis points toward 4.37%. According to Investing.com India, this divergence is particularly concerning as it occurred despite Micron delivering a quarter that should have reinforced every bullish AI narrative. The VIX spent much of the week pressing against 20 while the Mag 7 fell between 3% and 8%, indicating underlying market stress. The problem lies in simple arithmetic - when more than 30% of an index falls more than 5%, even strong performance from the remaining 490 stocks cannot prevent the benchmark from declining. As Investing.com India reports, the S&P 500 market-weighted stock price index peaked at a record high of 7,609.78 on June 2 and fell 3.4% through Friday's close, while the equal-weighted index remained unchanged over the same period.
Investors are experiencing AI Fatigue as they question whether hyperscalers' massive spending on AI infrastructure will ever pay off. According to Investing.com India, they see token prices falling, suggesting there might already be excess compute capacity, and worry about new technologies rapidly making current ones obsolete in a process known as 'creative destruction'. Companies are experiencing 'token budget hangover,' where agentic AI usage is blowing up forecasted budgets. Microsoft is eyeing DeepSeek as a hyper-cheap, optional alternative to expensive OpenAI and Anthropic models currently powering its enterprise agent tool, Copilot Cowork. To cut its own soaring internal AI token usage costs, Microsoft issued a firm June 30, 2026 cutoff deadline for its Experiences + Devices division to stop using Anthropic's Claude Code. The Mag-7, which includes the biggest hyperscalers, hasn't been so magnificent in June - the MAGS ETF peaked at a record high on May 26 and fell 12.9% through Friday's close, while the XMAG ETF is up 13.7%. Within the Information Technology sector, there has been rotation even within the semiconductor space - while the MAGS and IGV ETFs swooned during June, the SOXX semiconductor ETF continued to soar to a new record high on June 22.
According to a recent Wall Street Journal article, S&P 500 earnings growth is expected to top 20% for a second consecutive quarter, though this growth is not widespread across all sectors. The predominant bulk of earnings growth is being driven by semiconductor makers and AI infrastructure companies, creating a concentrated pattern of corporate performance. Morgan Stanley analyst Todd Castagno questions how long some tech companies can continue growing earnings at such historical rates, describing the current period as 'a golden window where everybody looks good'. The concentrated nature of this growth raises important questions about the sustainability of current market dynamics, with the market beginning to distinguish between companies collecting the AI scarcity premium and those financing the infrastructure race. As Investing.com India reports, FEMO (Fabulous Earnings Momentum) may be starting to lose a wee bit of its mojo - S&P 500 forward earnings edged up to another record high during the week of June 25 as both 2026 and 2027 analysts' consensus earnings expectations edged down, though they are still forecasting y/y growth rates in the low-to-mid 20s.
The semiconductor sector is experiencing significant growth driven by AI demand. SK Hynix, a Korean memory chip maker, announced plans for a $29.4 billion US listing, which would be the biggest American Depositary Receipt offering in history, surpassing Alibaba's $25 billion 2014 debut*, with the offering expected to begin trading on July 10. SK Hynix shares have already surged 300% in 2026 and more than 800% over the last full year, driven by strong HBM demand for AI applications. This semiconductor boom reflects the broader AI infrastructure investment cycle, with companies like Nvidia benefiting from immediate revenue recognition while downstream hyperscalers defer costs through capitalization accounting. Memory, HBM, networking and selected infrastructure suppliers sit on the right side of a bottleneck that everyone else is racing to secure, with scarcity transferring pricing power toward the owners of critical components.
According to the Wall Street Journal report, there are significant, growing expenses that will hit hyperscalers' income statements in the future. The question remains whether revenue can expand quickly enough to offset the coming bulge in expenses. As Zion Research Group founder David Zion notes, analyst depreciation estimates for each of these companies are all over the place, with consensus depreciation and amortization estimates potentially systematically understated. This makes it difficult to predict margins for 2028 and beyond, as the current accounting structure masks the true cost of AI infrastructure investments. The market is beginning to separate the AI ecosystem into those collecting the scarcity premium and those paying for it. The hyperscalers, device makers and enterprise users are the cheque writers - funding the compute, memory, power, cooling, networking and data-centre buildout while investors become more demanding about when the payoff arrives. If suppliers continue to outperform while hyperscalers lag, investors will increasingly treat the gap as a judgment on the economics of the AI spend.
