Manish Kabra, Société Générale's Chief US Equity Strategist, warns investors against over-allocating to tech stocks right now, despite their recent resurgence. He indicates that a safer investment opportunity will likely emerge around early 2027. Kabra's cautious stance stems from two key metrics he's monitoring among 'hyperscalers' – large tech companies aggressively pursuing AI ambitions. These companies include Amazon, Alphabet, Meta, Microsoft, Alibaba, Tencent, Oracle, Baidu, IBM, CoreWeave, Nebius, Salesforce, China Mobile, China Telecom, and China Unicom.
The first critical metric is free cash flow levels. Kabra notes that hyperscalers' free cash flow has been consistently declining since early 2024, as they pour substantial capital into AI infrastructure development. He projects that these firms' aggregate free cash flow will turn negative by the end of 2026 before recovering to positive territory in the first quarter of 2027. Kabra emphasizes that this positive inflection in free cash flow is essential for a significantly bullish outlook on the tech sector.
The second metric Kabra is closely watching is the hyperscalers' capital expenditure (capex) to sales ratio. These companies have been heavily investing in data centers for several years, with Meta, Amazon, Alphabet, and Microsoft alone anticipated to spend approximately $600 billion on AI in 2026, nearly doubling their 2025 expenditures. Kabra expects to see more concrete progress on monetizing these AI investments, and believes this will become evident in Q1 2027.
Given his reservations about the tech sector for the remainder of 2026, Kabra suggests that the S&P 500 will struggle to surpass 7,000, as tech stocks constitute roughly 32% of the cap-weighted index. He instead recommends considering the equal-weighted S&P 500 for broad market exposure, as it offers greater exposure to sectors like materials, industrials, and utilities, which are expected to benefit from increasing power demand in the US. Notably, demand for hyperscaler debt has recently waned, with cover ratios for hyperscaler bonds decreasing from 5x in February to 2x in July 2026, according to Apollo Global Management. Morgan Stanley estimates AI-related global debt issuance reached $236 billion as of May 2026, a fourfold increase year-over-year, and forecasts this to climb to $570 billion by the end of 2026.
