Nvidia posts $26 billion Q1 revenue amid record AI GPU demand surge
Gaming segment performs well too.
Nvidia on Wednesday reported its financial results for the first quarter of its fiscal year 2025, which ended on April 28, 2024. Due to overwhelming demand for its processors for AI and HPC workloads, the company recorded its best quarter ever and provided an even stronger outlook for its second quarter.
Nvidia reported a record quarterly revenue of $26.0 billion, up 18% from Q4 and up 262% year-over-year (YoY). Nvidia's GAAP net income for Q1 FY 2025 was $14.881 billion, up 21% from $12.285 billion in the previous quarter and up a whopping 628% from $2.043 billion in the same quarter a year before. The company's gross margin reached a phenomenal 78.4%, up from 64.6% in Q1 FY2024.
Datacenter
Datacenter is now Nvidia's main source of revenue and income. The green company's datacenter revenue hit a record $22.563 billion (which includes networking), increasing 427% year-over-year and 23% sequentially. That happened even though the company could no longer sell its high-performance GPUs to China.
Compute revenue reached $19.4 billion, up 478% year-over-year and 29% sequentially, driven by higher shipments of Hopper GPUs for AI applications. InfiniBand networking revenue was $3.2 billion, up 242% year-over-year but down 5% sequentially due to supply timing. Strong growth was observed across all customer types, especially enterprise and consumer Internet companies. Large cloud providers represented mid-40% of datacenter revenue and drove significant growth with the deployment of Nvidia AI infrastructure.
"Our datacenter growth was fueled by strong and accelerating demand for generative AI training and inference on the Hopper platform," said Jensen Huang, founder and CEO of Nvidia. "Beyond cloud service providers, generative AI has expanded to consumer internet companies, and enterprise, sovereign AI, automotive and healthcare customers, creating multiple multibillion-dollar vertical markets."
Graphics
Nvidia's gaming revenue increased to $2.647 billion, up 18% year-over-year, but declined 8% sequentially, which is in line with seasonality as sales of standalone graphics processors (including those for laptops) are typically down in the first calendar quarter. The annual rise was driven by higher demand, perhaps as Nvidia launched its GeForce RTX 40 Super-series products.
Professional Visualization revenue grew to $427 million, 45% year-over-year but fell 8% sequentially. The annual increase resulted from higher sell-in to partners as channel inventory levels normalized, and the sequential decline was primarily attributed to desktop workstation GPUs.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Automotive, OEM, other
Automotive revenue increased 11% year-over-year and 17% sequentially to $329 million, which is the higest result in a couple of years. The YoY growth was primarily driven by self-driving platforms, while the sequential rise was fueled by AI Cockpit solutions and self-driving platforms.
Nvidia's sales to OEMs were $78 million, which is flat from $77 million in the same quarter a year ago, and a decline from $90 million in Q4 FY2024.
Q2 Guidance
For Q2 of its fiscal 2025, Nvidia projects revenue of $28 billion ± 2%, with GAAP and non-GAAP gross margins expected to be around 74.8% and 75.5%, respectively. Operating expenses are anticipated to be approximately $4.0 billion on a GAAP basis and $2.8 billion on a non-GAAP basis.
"We are poised for our next wave of growth," said Huang. "The Blackwell platform is in full production and forms the foundation for trillion-parameter-scale generative AI. Spectrum-X opens a brand-new market for us to bring large-scale AI to Ethernet-only data centers. And Nvidia NIM is our new software offering that delivers enterprise-grade, optimized generative AI to run on CUDA everywhere — from the cloud to on-prem data centers and RTX AI PCs — through our expansive network of ecosystem partners."
Nvidia also announced a ten-for-one stock split effective June 7, 2024, aimed at making stock ownership more accessible. Additionally, the company increased its quarterly cash dividend by 150% to $0.01 per share on a post-split basis, payable on June 28, 2024.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
AlskiOnTheWeb With so much of the revenue coming from data center and so little from gaming ... you'd think they would focus more on the data center side honestly. It's got to be a lot more lucrative in the long run.Reply -
renz496
nvidia still need their "gaming segment" to dominate the professional space. the entrance to nvidia CUDA ecosystem end up being cheap thanks to their gaming card. and right now nvidia can charge what gamer assume as absurd overprice with their gaming card thanks to semi pro that willing to buy nvidia GPU at higher price. and their numbers start eclipsing actual gamer that buy those GPU for gaming only.AlskiOnTheWeb said:With so much of the revenue coming from data center and so little from gaming ... you'd think they would focus more on the data center side honestly. It's got to be a lot more lucrative in the long run. -
Notton
But they are doing just that?AlskiOnTheWeb said:With so much of the revenue coming from data center and so little from gaming ... you'd think they would focus more on the data center side honestly. It's got to be a lot more lucrative in the long run.
5080 and 5090 are using GB203
This is a huge departure from before, where Ada Lovelace, Ampere, and Turing used 102, 103, etc.
GB100 is entirely reserved for AI.
If you look at historic Nvidia earnings, you'll notice that data center is growing exponentially.
It's kind of hard to judge its mid and long term performance. Is it going to burst, deflate, maintain, or grow even more? Not a good thing to throw all your eggs into.
Nvidia is not strapped for cash, and AMD and Intel are struggling with GPUs. It doesn't make sense to part with a market that they've already captured around 70~80% of.
The only area that is insufficient is TSMC capacity, but they aren't using the latest and greatest for gaming GPUs. Instead, they are using that for data centers.
They can also switch gaming GPUs back to Samsung, like they did with RTX 3000 series. -
edzieba
If that were the case, we would be expecting gaming revenue to be trending up along with DC revenue. Instead, it's up a little but mostly flat compared to the rocketship of DC revenue, and without the inflection in Q1-Q2 2024 that DC saw, indicating the two are not coupled by the same demand.renz496 said:nvidia still need their "gaming segment" to dominate the professional space. the entrance to nvidia CUDA ecosystem end up being cheap thanks to their gaming card. and right now nvidia can charge what gamer assume as absurd overprice with their gaming card thanks to semi pro that willing to buy nvidia GPU at higher price. and their numbers start eclipsing actual gamer that buy those GPU for gaming only. -
JarredWaltonGPU
This is why Nvidia has announced Blackwell B100/B200 and Grace Blackwell GB200 already. Those will take time to ramp up production, but Nvidia now knows it can sell basically as many as TSMC can produce. They're still on TSMC 4nm, though (4NP this time, not vanilla 4N). Consumer and professional Blackwell GPUs will probably use 4NP as well, but since that's not the most advanced node at TSMC, pricing and capacity should be decent.AlskiOnTheWeb said:With so much of the revenue coming from data center and so little from gaming ... you'd think they would focus more on the data center side honestly. It's got to be a lot more lucrative in the long run.
Just wait until next year when Nvidia announces the 3N or 2N successor to B200. LOL -
TerryLaze What black magic is that?! They increased their revenue by 260% their margins went up a little bit, but their income increased by 600% +Reply
How?!
Is that just weird calculation that tax people use?! -
renz496
production limit. most of nvidia 5nm capacity are being channeled towards professional card. nvidia knows gamer in general are buying less. there is no need to excessively increase production. semi pro are growing but nvidia will be appreciate more if those that buying multiple 4090 to get true pro cards like A5000/A6000 instead. but gaming revenue at 2.6 billion is already more than what nvidia can ask for. it could be worse like what happen during turing generation.edzieba said:If that were the case, we would be expecting gaming revenue to be trending up along with DC revenue. Instead, it's up a little but mostly flat compared to the rocketship of DC revenue, and without the inflection in Q1-Q2 2024 that DC saw, indicating the two are not coupled by the same demand. -
edzieba
If there was a 'production limit', and if the original claim that the majority of gaming cards were being purchased for 'semi pro' use was true, then we would expect to see the margins for the gaming card sector spike even for flat revenue, as prosumer use leans towards the highest end (and highest margin) SKUs closest to the professional card capabilities, whereas gaming sales trend towards the midrange higher-volume-lower-margin cards. We do not see this.renz496 said:production limit. most of nvidia 5nm capacity are being channeled towards professional card. nvidia knows gamer in general are buying less. there is no need to excessively increase production. semi pro are growing but nvidia will be appreciate more if those that buying multiple 4090 to get true pro cards like A5000/A6000 instead. but gaming revenue at 2.6 billion is already more than what nvidia can ask for. it could be worse like what happen during turing generation.
In addition, the main production limit for Nvidia was not die fabrication or PCBA production, but specifically the CoWoS packaging step for SKUs with HBM. As none of the consumer cards utilise HBM, this bottleneck would not be in effect. -
renz496
sorry when i said "production limit" i was meant to say nvidia limiting 40 series production. there are already reports about partner receiving reduced amount of supply of 40 series GPU. as for semi pro the interest not necessarily on the highest end only. some partner (i think it was PNY) build 4070 with blower cooler. in their marketing material they mention such cooler is the best option when the GPU is use in tandem inside the system. there is no gamer use multi GPU ever since nvidia completely ditch SLI support. and recently Galax releasing single slot 4060Ti specifically for 16GB model only. spec and performance wise those 4060Ti 16GB is similar to nvidia A4000 Ada. unless you need more specific support from nvidia that comes with true pro card getting GTX4060Ti at $500+ is still cheaper than A4000 that starts at $1200. in some place the A4000 Ada end up costing as high as $2000 due to demand.edzieba said:If there was a 'production limit', and if the original claim that the majority of gaming cards were being purchased for 'semi pro' use was true, then we would expect to see the margins for the gaming card sector spike even for flat revenue, as prosumer use leans towards the highest end (and highest margin) SKUs closest to the professional card capabilities, whereas gaming sales trend towards the midrange higher-volume-lower-margin cards. We do not see this.
In addition, the main production limit for Nvidia was not die fabrication or PCBA production, but specifically the CoWoS packaging step for SKUs with HBM. As none of the consumer cards utilise HBM, this bottleneck would not be in effect.