[News] NVIDIA Forecasts Q1 Sales to Rise 65%, Reinforcing AI Demand Amid DeepSeek Concerns

The AI frenzy is not over yet, as U.S. chip giant NVIDIA beats Q4 estimates and gives a strong outlook for the current quarter. The company forecasts Q1 sales to hit $43 billion, up 65% YoY. As highlighted by Reuters, this signals strong AI demand from Microsoft, Amazon, and other tech giants, easing concerns sparked by DeepSeek’s low-cost model launch last month.

Meanwhile, NVIDIA projects a Q1 gross margin of 71%, down from 73.5% last quarter. Reuters, citing CFO Colette Kress, points out that margins should return to the mid-70% range later this fiscal year as Blackwell chip production scales up, cutting costs.

Blackwell Sales Hit $11 billion in Q4

As per Reuters, CEO Jensen Huang remains upbeat about the AI boom, calling AI’s progress “light speed” and Blackwell demand “amazing.” A report from yahoo!finance notes that NVIDIA’s Blackwell AI chips, now in full production, generated $11 billion in Q4 revenue and marks the company’s fastest product ramp.

NVIDIA beat Wall Street expectations in its Q4 earnings, with revenue up 78% to $39.3 billion. Its FY25 revenue jumped 114% to $130.5 billion, as per its press release. Its FY25 GAAP earnings per diluted share was $2.94, up 147% from a year ago.

Large cloud providers drove Blackwell sales, accounting for about 50% of NVIDIA’s Data Center revenue in the quarter, according to CNBC. The momentum is set to continue, as the company expects a significant ramp of sales of Blackwell in the first quarter, the report adds.

NVIDIA’s data center unit, which includes Blackwell and Hopper chip sales, now makes up 91% of total revenue, up from 83% last year, as per CNBC.

DeepSeek Drives Demand for AI Inference Chips

Notably, NVIDIA also dismissed concerns that efficient models like DeepSeek’s R1 could reduce its chip demand. The CNBC report, citing CFO Kress, notes that advanced AI methods may need up to 100 times more chips, adding that NVIDIA’s new Blackwell chips will focus on AI inference, not just training. According to NVIDIA, B200 could have 25x better inference throughput.

TrendForce highlights that DeepSeek’s influence will drive CSPs toward lower-cost proprietary ASIC solutions, shifting focus from AI training to AI inference. This shift is expected to gradually increase the share of AI inference servers to nearly 50%.

According to Reuters, Chinese companies, like Tencent, Alibaba, and ByteDance, are ramping up orders for NVIDIA’s H20 AI chips due to the growing demand for DeepSeek’s low-cost AI models.

Read more

• [News] TSMC Reportedly Sees CoWoS Order Surge, with NVIDIA Securing 70% of 2025 CoWoS-L Capacity
• [News] DeepSeek Adoption Reportedly Fuels NVIDIA H20 Orders, Led by Tencent, ByteDance, Alibaba

(Photo credit: NVIDIA)