Nvidia's growth slows to a mere 122 percent but it’s still topping expectations
- Reference: 1724905750
- News link: https://www.theregister.co.uk/2024/08/29/nvidias_growth_slows/
- Source link:
The GPU giant today revealed Q2 2025 revenue of $30.04 billion – up 122 percent for the year. But it's also rather less than the 262 percent, 265 percent, and 206 percent year-over-year growth the chip shop reported in Q1 2025, Q4 2024, and Q3 2024. Quarter-to-quarter growth has also slowed: for this quarter it was 15 percent, compared to 17 percent, 22 percent, and 34 percent in preceding quarters.
Nvidia isn't worried in the slightest as it predicted $28 billion revenue, plus or minus two percent.
[1]
Datacenter kit dominated, contributing $26.3 billion revenue, up 16 percent from Q1 and 154 percent from a year ago. Cloud service providers accounted for 45 percent of that haul, while more than 50 percent came from consumer internet and enterprise customers.
[2]
[3]
Networking accounted for $3.7 billion of the datacenter total. With an annual run rate of over $14 billion, Nvidia is therefore the second-biggest datacenter networking vendor – and bigger than Arista and Juniper combined.
Investors were told that H200 series accelerators, which use Nvidia's Hopper architecture, are starting to ramp, and big clouds are lining up to buy them.
[4]
CEO Jensen Huang said that's because "… if you just look at the world's cloud service providers, the amount of GPU capacity they have available, it's basically none. And the reason for that is because they're either being deployed internally for accelerating their own workloads" or renting them to startups.
"If you have a choice between building CPU infrastructure right now for business or Hopper infrastructure for business right now, that decision is relatively clear," Huang explained, "So, I think people are just clamoring to transition the $1 trillion of established installed infrastructure to a modern infrastructure and Hopper's state-of-the-art."
It's a good thing, then, that supply and availability of Hopper products has improved.
[5]
The acceleration champ also revealed it's shipped samples of its forthcoming Blackwell product range – its next-gen tech – and predicted it will bring more billions to the balance sheet this year.
[6]Delays? We're still shipping 'small quantities' of Nvidia's GB200 in Q4, Foxconn insists
[7]Huawei's Ascend 910 launches this October to challenge Nvidia's H100
[8]Tenstorrent's Blackhole chips boast 768 RISC-V cores and almost as many FLOPS
[9]Intel finally has a new GPU – for cars
Despite US bans on sales of advanced tech to China, revenue from the Middle Kingdom rose. In a [10]regulatory filing [PDF],Nvidia noted "Our datacenter revenue in China grew sequentially in the second quarter of fiscal year 2025 and is a significant contributor to our datacenter revenue" – but still a smaller contributor than it was before sanctions were imposed.
The filing also notes that China's [11]new energy regulations – which set standards for compute performance per watt and per memory bandwidth of accelerators used in new and renovated datacenters – could change in ways that mean Nvidia would not be able to create products that would be allowed in the Middle Kingdom.
Another nugget in the filing is a colossal sales spike in Singapore, which accounted for $5.6 billion revenue – up from $1 billion last quarter. Nvidia's posted similar oddities before, and in this quarter felt the need to inform investors that "most shipments associated with Singapore revenue were to locations other than Singapore and shipments to Singapore were insignificant."
Huang used the earnings call to deliver his usual burst of AI optimism, with associated boasting that Nvidia alone can make it happen. He predicted $32.5 billion of revenue for its next quarter – year-over-year growth of a mere 79.5 percent.
Our sibling publication The Next Platform has covered remarks by CFO Colette Kress [12]here .
Investors didn't like what they heard: Nvidia's shares dropped from around $126 apiece to the $116 range in after hours trading. ®
Get our [13]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/cloudinfrastructuremonth&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZtBGxed2hNwme6BLhQTYJgAAAM4&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/cloudinfrastructuremonth&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZtBGxed2hNwme6BLhQTYJgAAAM4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/cloudinfrastructuremonth&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZtBGxed2hNwme6BLhQTYJgAAAM4&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/cloudinfrastructuremonth&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZtBGxed2hNwme6BLhQTYJgAAAM4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/cloudinfrastructuremonth&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZtBGxed2hNwme6BLhQTYJgAAAM4&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2024/08/14/nvidia_foxconn_blackwell/
[7] https://www.theregister.com/2024/08/13/huaweis_ascend_910_launches_this/
[8] https://www.theregister.com/2024/08/27/tenstorrent_ai_blackhole/
[9] https://www.theregister.com/2024/08/08/intel_a760a_gpu/
[10] https://d18rn0p25nwr6d.cloudfront.net/CIK-0001045810/78501ce3-7816-4c4d-8688-53dd140df456.pdf
[11] https://www.theregister.com/2024/08/13/china_green_policy_tech_elements/
[12] https://www.nextplatform.com/2024/08/28/in-depth-post-earnings-review-with-nvidia-cfo-colette-kress/
[13] https://whitepapers.theregister.com/
Re: What ?
Possibly or alternatively shipping to Singapore and then the customer ships somewhere else (eg China,Hong Kong, Iran,North Korea,Russia) or company setup in Singapore shipping somewhere else and then that customer ships to destination country.
It's possible although less likely that suddenly Singapore has grown as a place for shell companies or offshore subsidiaries for GPUs that much for non sanctions busting reasons. I know it has been popular in the past for tax minimisation and avoidance.
I wonder how many Chinese companies simply redirected their orders to subsidiaries data centres in non sanctioned countries , or to dedicated cloud resources they rent. Ie the Chinese companies effectively control the compute it's just under a subsidiary now or in someone else's data centre or cloud.
China has a mix of both onshore and offshore strategies with the ideal mix depending upon China future plans and timelines eg invading Taiwan and it's expectation of the risk of further actions by the west.
Ideally China would like sovereign compute with their own designs and manufacturing capabilities but while they are pursuing that and developing it, they still want western infrastructure whether it's sanctioned or not.
China also can also fund dedicated compute infrastructure overseas in different offshore locations depending upon where it expects sanctions and repercussions to bite eg severing fibre optic cables in event of invasion of Taiwan. If China expected to invade one or more countries one strategy would be to build up compute infrastructure and capacity in those countries and then take it over.
Overvalued?
I don't think it's a good time to buy shares in Nvidia. The crypto boom is over, and AI optimism is starting to creak. There's no other compute-hungry fad in sight. What then?
Re: Overvalued?
An interesting question to ask is who is actually using the ML fad?
Bunch of IT cost and personnel soaked up for months (years?) in many businesses. What you actually get on the end is a toy that your staff play with for half an hour, ask a few questions where responses are completely wrong... Followed by the observation "that's a bag' o shite"; and have no reason to look at it ever again.
Applied statistics is a very potent tool used in the right locations, but trying to use it as some piss-poor text parser that consumes ludicrous amounts of power to run does not stack up.
In the meantime a stack of expensive GPU's in server banks depreciate not doing anything particularly useful.
I don't like the idea of legislation to stop the waste; but perhaps, there is a case for it here.
.. cause & effect
Drop are mainly coz of China massively moved Ai on theirs supercomputers, and ofc there is a new Ascend 910C Ai card by Huawei, choice of Baidu, ByteDance & China Mobile in a fitst row.
Nvidia Ai ballooning & exaggerated worth will eventually burst, mainly because their projection are based on the good sell in China's market.
What ?
" most shipments associated with Singapore revenue were to locations other than Singapore and shipments to Singapore were insignificant "
Well then why post that revenue under Singapore ?
Is it because you were actually shipping to China when you're not supposed to ?