Qualcomm Says Data Center Chips to Produce ‘Billions’ in ’27

Bloomberg analyst Mandeep Singh previewed Micron's earnings and framed the broader AI chip market as one where suppliers may retain pricing power, similar to oil producers influencing energy prices. He suggested chipmakers are unlikely to materially expand AI memory capacity while demand remains strong. The commentary is more strategic than event-driven and does not include specific earnings figures or guidance.

Analysis

The key market implication is not just that AI demand is strong, but that pricing power in memory can become a self-reinforcing bottleneck. If suppliers keep capacity disciplined, hyperscalers will be forced to choose between paying up for continuity or redesigning around tighter allocations, which pushes bargaining power upstream and extends the margin cycle for the memory complex. The second-order winner is likely equipment and materials vendors only if there is a lagged capex response; otherwise, the near-term cash flow gets captured by the incumbents rather than reinvested.

The larger risk is that the market is assuming supply scarcity is durable, when memory is historically the most reflexive part of semis. A sharp step-up in AI capex over the next 2-3 quarters could trigger a delayed supply response by contract manufacturers and tool vendors, compressing margins 12-18 months out even if spot pricing stays firm in the interim. That means the trade is better expressed as a medium-term earnings momentum story than a secular perpetual rerating.

A contrarian view is that the current debate may be underestimating substitution risk. If memory becomes structurally expensive, customers will optimize around fewer, faster deployments, shift workloads to more memory-efficient architectures, or delay some inference capacity, which can eventually slow the demand curve for the very products benefiting today. In other words, sustained scarcity may maximize near-term profits but also incentivize design changes that cap the long-run oligopoly premium.

From a trading standpoint, the cleanest setup is to own the names with the strongest near-term operating leverage and avoid paying peak-multiple valuations for the broader AI basket. The best risk/reward is often a pair between memory beneficiaries and more fully valued AI enablers whose upside already reflects uninterrupted capex acceleration. Event risk is concentrated around earnings guidance and any commentary on capacity additions, inventory normalization, or pricing assumptions over the next 1-2 reporting cycles.

Key Decisions for Investors

• Go long a basket of memory leaders into earnings strength over the next 2-6 weeks, but only on pullbacks; the setup favors cash-flow capture over valuation expansion, with the main risk being a guidance miss or softer pricing commentary.
• Pair trade: long memory suppliers / short high-multiple AI infrastructure names over 3-6 months; this captures the likelihood that scarcity economics accrue to the bottleneck rather than the downstream ecosystem.
• Buy downside protection on the broad AI hardware complex for the next 1-2 quarters; if hyperscalers signal capex discipline or design changes, the de-rating can be swift even if semiconductor demand remains healthy.
• If Micron-guidance analogues point to continued capacity restraint, add to the trade only after earnings and conference call confirmation; the risk/reward improves when management explicitly prioritizes margins over unit growth.
• Avoid chasing any breakout higher in the memory group without evidence of supply response timing; the historic failure mode is a delayed oversupply cycle that appears 12-18 months after peak enthusiasm.