Huang says Nvidia’s chip sales will double next year

Nvidia CEO Jensen Huang said the company expects to sell twice as many chips next year as this year, reinforcing expectations for sustained AI-driven demand over the next six quarters. Nvidia has also projected roughly 70% growth in the fiscal year ending January 2028, reaching about $673 billion. Huang attributed the outlook to broad global investment in AI across industries and economies, while noting that AI products should be withheld if safety standards are not met.

Analysis

The key investment issue is not unit demand but whether NVDA can sustain revenue growth while preserving its extraordinary accelerator-system economics. A doubling of chip volumes would require hyperscalers and sovereign buyers to fund both compute and increasingly scarce power, networking and data-center buildouts; the bottleneck therefore shifts from GPU availability toward electricity interconnects, rack integration and customer ROI. This favors adjacent suppliers with content-per-rack exposure—AVGO, ANET, VRT and ETN—while raising the risk that NVDA's growth converts into lower-margin mix if supply broadens faster than premium compute demand.

Near-term, the statement should reinforce estimates and support NVDA sentiment over the next 1-3 months, but it is not independently verifiable guidance because chip units are not a disclosed KPI and product configurations differ materially. The relevant earnings catalyst is whether data-center revenue, gross margin and receivables remain consistent with demand being end-customer led rather than inventory accumulation by a concentrated cloud customer base. A gross-margin reset, a sequential slowdown in networking attach, or weaker capex commentary from GOOG/MSFT/AMZN/META would matter more than further unit-demand rhetoric.

Consensus may be underestimating the second-order constraint: AI capex can remain robust while the value pool broadens away from NVDA toward power and physical infrastructure. Conversely, a sustained decline in rates would ease financing constraints for data-center projects and extend the capex cycle, making a simplistic NVDA short premature. Over 6-18 months, competitive risk is more likely to emerge through customer-built ASICs and workload-specific inference economics than through a sudden GPU substitution event.

Key Decisions for Investors

• Maintain an NVDA core long into the next earnings print, but fund part of the exposure with a 3-6 month call spread rather than outright incremental shares; upside requires another estimate revision cycle, while a data-center gross-margin or backlog-quality disappointment can drive sharp multiple compression.
• Add a basket long of VRT and ETN against a smaller NVDA short only after a strong NVDA-led rally: the trade targets the shift from compute scarcity to power/cooling deployment over 6-18 months. Falsify if hyperscaler capex guidance decelerates broadly, rather than merely rotating within the AI supply chain.
• Prefer ANET over GOOG as a liquid AI-adjacency expression for the next 1-3 months; accelerating cluster scale raises network intensity, whereas GOOG must translate capex into cloud/advertising monetization and faces a less direct earnings transmission.
• Set an earnings watch item—not a trade trigger—for NVDA: reduce exposure if sequential data-center growth slows materially without an offsetting gross-margin improvement, or if major cloud customers signal lower AI infrastructure spending. Those outcomes would challenge both volume assumptions and the sector's elevated capex-duration multiple.