Jetmek_03052
9時間前
Nvidia Stock Has Almost Recovered From DeepSeek Rout. There's More Good News. -- Barrons.com
February 18, 2025
04:43 PM ETPublished February 18, 2025 04:43 PM Eastern TimeDow Jones Newswires
Callum Keown, Adam Clark, and Tae Kim
Nvidia stock moved within touching distance of recouping all its losses from the DeepSeek market rout on Tuesday
The company's shares rose 0.4% to $139.40, closing in on the $142 level where it was trading before the Chinese start-up's apparently low-cost artificial-intelligence model roiled tech stocks last month.
Nvidia stock tumbled 17% on Jan. 27, reducing the AI chip maker's market capitalization by close to $600 billion, the biggest one-day loss ever by a U.S. company.
But just three weeks later, it has almost fully recovered. Investors who bought the dip can afford a smile as they look toward the company's earnings next week.
There's more good news. The South Korean government announced plans to acquire 10,000 graphics-processing units as it looks to build a national AI computing center. That includes Nvidia's H100 and H200 GPUs, South Korea's Yonhap News Agency reported Monday.
Hargreaves Lansdown analyst Matt Britzman said it is another sign that "Nvidia's demand extends well beyond the giant U.S. tech companies."
He added: "We've now seen several countries express an appetite for building their own computing clusters, with the enormous U.S. Stargate project grabbing the most headlines. This is supportive to the Nvidia investment case, and presents a relatively new and scalable demand avenue for its market leading chips."
One concern for Nvidia has been whether rental prices for its existing chips will hold up as companies and startups compete to buy its newest Blackwell hardware. UBS analyzed the latest data from the largest cloud-computing providers and found pricing is stable for renting Nvidia's Hopper H100 chips, suggesting there is still a lack of capacity for GPU computing.
"Buoyant cloud GPU pricing affirms strong demand environment," wrote UBS analyst Timothy Arcuri in a research note Monday. "Looking ahead, given the demand backdrop we would not be surprised if pricing for Hopper instances remains resilient."
Arcuri has a Buy rating on Nvidia stock with a $185 target price.
Write to Callum Keown at callum.keown@dowjones.com and Adam Clark at adam.clark@barrons.com
This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.
IanFromSI
13時間前
Here’s a three paragraph taste from that article and Barrons.
“Sell-side ratings are mostly useless,” wrote Trivariate Research founder, and former Morgan Stanley chief U.S. equity strategist Adam Parker in a Friday report. Sell-side refers to analyst firms on Wall Street that “sell” research to the buy-side—mainly large institutions that manage trillions of dollars.
That’s quite a statement from Parker. To justify it, he looked at stock returns for the past 25 years and found that the best returns were concentrated in the least-loved stocks. Still, that doesn’t mean investors should pile into the least-loved stocks and wait, he added. It just means that buying up the best-loved stocks on Wall Street isn’t a great idea.
One reason investors can’t overreact to any one signal or research observation is that things change—a lot. There are regulatory changes that upend Wall Street, such as Reg FD, or regulation fair disclosure, implemented in the aftermath of the dot.com bubble that stopped companies from selectively disclosing information to analysts and large shareholders. There are also hosts of investors looking for an edge who will arbitrage gains from any new informational signal.
JJ8
13時間前
Prediction: Nvidia Stock Is Going to Drop After Feb. 26
Timothy Green, The Motley Fool
Tue, February 18, 2025 at 5:15 AM PST
Tech giants are dumping mountains of cash into artificial intelligence (AI) data centers. Microsoft plans to spend $80 billion this year to expand its AI capacity. Amazon is boosting its total capital spending, spread across its retail and cloud businesses, to $100 billion to accelerate its AI efforts. And Meta Platforms will pour $65 billion into its data centers to fuel its AI ambitions.
On the surface, this all sounds great for Nvidia(NASDAQ: NVDA) stock. The company dominates the market for powerful AI accelerators, and second place AMD has already put forth a disappointing forecast for its own AI chips sales this year. Nvidia is going to scoop up the lion's share of the spending that goes toward AI accelerators.
Where to invest $1,000 right now? Our analyst team just revealed what they believe are the 10 best stocks to buy right now. Learn More »
Trees don't grow to the sky
Forecasts for AI accelerator sales paint a rosy picture for Nvidia. AMD, for example, once predicted that AI accelerators would generate $500 billion in industrywide revenue in 2028.
However, investors need to ask an important follow-up question: How can that spending possibly be justified?
If companies are going to spend half-a-trillion dollars a year on AI accelerators, not to mention many billions more on other data center gear, those investments will need to pay off in the form of new sources of revenue or cost savings.
Is that realistic? What will generate that new revenue?
Nvidia's valuation is built on optimism -- namely, optimism that its revenue and profit can continue to grow at blistering rates for years. The stock currently trades for more than 40 times expected earnings for fiscal 2025.
This is a company that's already worth more than $3 trillion and generated $20 billion in adjusted net income last quarter. The market for AI accelerators must continue to grow rapidly for Nvidia's stock price to make any sense.
It's not that investors are discounting the possibility that demand for AI accelerators flatlines -- it's that investors are discounting the possibility that demand collapses. What will happen to Nvidia stock if, after Microsoft, Amazon, and Meta hurl hundreds of billions of dollars into AI data centers, those companies fail to generate a reasonable return on investment?
The AI investment boom feels like a massive case of FOMO (fear of missing out). Tech giants are terrified of being left behind, so they're throwing caution to the wind.
There's genuinely an enormous amount of demand for AI computing capacity right now, but how much of that is experimentation? In other words, how much of that is companies trying out AI to see if it makes financial sense? When some of those experiments don't pan out, what will happen to demand?
Then there's DeepSeek, the Chinese company that trained an AI model that can compete with the best AI models U.S. companies have to offer for a fraction of the price. If training top-tier AI models no longer requires megaclusters of AI accelerators, what will happen to demand?
This might be peak Nvidia
I'm calling it: Peak Nvidia is almost here, and it's close to the point where the stories and predictions keeping the AI boom afloat start to fall apart. It's not that AI isn't an impressive and useful technology. Like the internet, it's revolutionary.
But also like the internet 25 years ago, it's creating expectations that appear untethered from reality. The internet changed the world but also ruined plenty of investors along the way.
I could be very wrong about this. It's certainly possible that the hundreds of billions of dollars being spent on AI infrastructure will make financial sense in the end and demand for AI accelerators will continue to grow far into the future. Maybe Nvidia will put out a stellar forecast that drives the stock to new highs. Anything is possible.
Still, I don't think I'd want to be an Nvidia shareholder on Feb. 26 when the company reports its latest results. Expectations are sky high, and any hint of trouble could send the stock plunging.
Don’t miss this second chance at a potentially lucrative opportunity
Ever feel like you missed the boat in buying the most successful stocks? Then you’ll want to hear this.
On rare occasions, our expert team of analysts issues a “Double Down” stock recommendation for companies that they think are about to pop. If you’re worried you’ve already missed your chance to invest, now is the best time to buy before it’s too late.
4retire
20時間前
TAKE
Nvidia Stock Has Almost Recovered From DeepSeek Rout. There’s More Good News.
Investors who bought the dip can afford a smug smile as they look toward the company’s earnings next week.
Follow Barron's in Apple News
N?vidia stock rose early Tuesday as it moved within touching distance of recouping all its losses from the DeepSeek market rout.
The shares rose 0.8% to $139.97 ahead of the open, closing in on the $142 level at which it was trading before the Chinese start-up’s apparently low-cost artificial-intelligence model roiled tech stocks last month.
Nvidia stock tumbled 17% on Jan. 27 as the AI chip maker’s market capitalization fell by close to $600 billion, the biggest one-day loss ever by a U.S. company.
But just three weeks later it’s almost fully recovered, and investors who bought the dip can afford a smile as they look toward the company’s earnings next week.
There’s more good news. The South Korean government announced plans to acquire 10,000 graphics-processing units (GPUs) as it looks to build a national AI computing center. That includes Nvidia’s H100 and H200 GPUs, South Korea’s Yonhap News Agency reported Monday.
Hargreaves Lansdown analyst Matt Britzman said it’s another sign that “Nvidia’s demand extends well beyond the giant U.S. tech companies.”
He added: “We’ve now seen several countries express an appetite for building their own computing clusters, with the massive U.S. Stargate project grabbing the most headlines. This is supportive to the Nvidia investment case, and presents a relatively new and scalable demand avenue for its market leading chips.”
One concern for Nvidia has been whether prices for its existing chips will hold up as companies and governments compete to buy its newest Blackwell hardware. The latest data from UBS suggest pricing is stable for Nvidia’s H100 chips and even its older A100 processors.
“Pricing for most GPU instances has remained steady (and in most cases is tending, if anything, higher)—a bullish read-through that supply remains the growth limiter for Nvidia, rather than demand,” wrote UBS analyst Timothy Arcuri in a research note.
Arcuri has a Buy rating on Nvidia stock with a $185 target price.
eastunder
21時間前
DeepSeek Did Nvidia a Favor
https://www.msn.com/en-us/money/topstocks/deepseek-did-nvidia-a-favor/ar-AA1zcm4B?ocid=BingNewsSerp
DeepSeek hasn’t sunk Nvidia’s prospects for a big year. But more-grounded hopes should still prove advantageous to the artificial-intelligence giant as it navigates a tricky road ahead.
The panic sparked by the Chinese AI startup’s revelations last month cost Nvidia nearly $750 billion in market cap in a little over a week’s time. The stock has recovered some since, but is still down about 6% as of Friday’s close compared with the Nasdaq’s flat performance over the same period.
DeepSeek’s claims to have produced an advanced AI model at a relatively low computing cost raised the prospect that demand could wither for Nvidia’s expensive systems that have been considered must-haves for any company looking to build up generative AI capabilities.
The recent spate of earnings reports from Nvidia’s largest customers strongly suggest otherwise. Amazon, Meta Platforms and Google-parent Alphabet all projected sharp increases in capital spending for this year—on top of a surge last year. All three also touted their relationships with Nvidia in their earnings calls.
Amazon Chief Executive Andy Jassy said the “deep partnership” between the two will last “for as long as we can see into the future.” Amazon expects capital spending to top $100 billion in 2025, with most of it going to AI infrastructure.
That bodes well for Nvidia’s own quarterly report on Feb. 26. The data-center segment that includes most of the company’s AI chips and services is expected to show revenue more than doubling to $113 billion for the fiscal year ended January.
But there is also growing worry that a transition to Nvidia’s new Blackwell chip family could cause a short-term hiccup in growth. Data-center revenue in the fiscal fourth quarter is expected to have risen by only $2.6 billion from the period ended October, according to estimates from FactSet.
That would be a sharp deceleration from the $4.5 billion in data-center sales added in the previous quarter—and the lowest sequential gain for that segment since AI demand started booming in early 2023.
Nvidia CEO Jensen Huang said on the company’s last earnings call in November that Blackwell production “is in full steam.” But many analysts don’t expect those Blackwell shipments to hit high volume until the second half of the new fiscal year, which could mean a disappointing forecast from Nvidia for its April-ending quarter. Mark Lipacis of Evercore ISI said an “air pocket” in shipments is the biggest risk to his positive outlook on Nvidia’s stock heading into the report.
Nvidia’s clipped valuation could thus be a blessing in disguise, if it helps ground expectations a bit ahead of what could be a mixed report. The stock is now trading a little under 32 times projected earnings for this year, compared with a forward earnings multiple of nearly 40 times six months ago.
Nvidia is also now cheaper by that measure than most of its megacap tech peers despite superior growth projections, with revenue projected to rise 53% this year compared with 12.2% average growth expected for Apple, Microsoft and the other tech giants fetching market caps of more than $1 trillion each, according to FactSet data.
And most of those companies are still planning to shovel many billions more Nvidia’s way. DeepSeek won’t be making Nvidia come up short just yet.
rolvram
22時間前
Grok 3 is officially the Best AI chatbot in the world!
This is HUGE News.
But here's where it gets interesting.
@xai was able to do this by betting that "scaling laws" would hold.
What does that mean?
It means that if you have a bigger unified AI supercomputer to train with, you can create a smarter base AI system.
This fact is TERRIFYING for every other company and nation in the world working on frontier AI models because @xai already has the world's biggest unified training cluster, and it's growing much faster than anyone can keep up with.
It only took 4 months for xAI to build a working, unified training cluster of over 100k @nvidia H100s.
The previous largest was ~30k H100s, so they more than tripled the world's best.
But xAI didn't stop there.
Within 3 months after completion of the first 100k cluster, they were able double it to 200k (or an equivalent amount of higher performance GPUs).
This lead in training cluster size looks unassailable.
@elonmusk's unparalleled execution makes him one of Jensen Huang's favorite customers, which means xAI will get first priority when NVIDIA starts shipping their next generation of bigger, better, and faster GPUs.
And that's not all.
xAI is also MUCH FASTER at getting new GPUs added to their cluster once they've been received than any other company.
Between their existing lead, their ability to acquire next generation GPUs before competitors, and their lightning fast speed in putting those new GPUs to work in training, how is anyone supposed to compete with xAI?
doc2016
2日前
the idea of a virtual pbx keeping track and a registry of quantum entangled and their states would seem to yield information about how best to maintain quantum coherence and uptime/qos? all of that could be put toward the yield of logical qubits from physical qubits being optimized? including the settings/programming of such and their photonic management? for instance what is the optimal sampling of the qubits, timing, energy wavelengths, magnetic fields, temperature, qubit types, silicon dioxide insulation use, etc.
intel previously announce a photonics research place in ? texas. now broadcom and others are said to be interested in the intel photonics r and d division.
such a communications capabilities would make cuda-q better going forward?
4retire
3日前
Broadcom, TSMC eye possible Intel deals to split storied chipmaker, WSJ reports
PUBLISHED SUN, FEB 16 20257:03 AM EST
Reuters
WATCH LIVE
In this article
AVGO
-0.43 (-0.18%)
After Hours
2330-TW
-30.00 (-2.75%)
INTC
+0.01 (+0.04%)
After Hours
Follow your favorite stocks
CREATE FREE ACCOUNT
The logo of Intel is seen during Computex 2024 in Taipei on June 4, 2024. (Photo by I-Hwa CHENG / AFP) (Photo by I-HWA CHENG/AFP via Getty Images)
The logo of Intel is seen during Computex 2024 in Taipei, Taiwan, June 4, 2024.
I-hwa Cheng | Afp | Getty Images
Intel’s
rivals Taiwan Semiconductor Manufacturing Co.
and Broadcom
are each eyeing potential deals that would break the U.S. chipmaking icon in two, the Wall Street Journal reported on Saturday, citing people familiar with the matter.
Broadcom has been closely examining Intel’s chip design and marketing business, the Journal reported, adding that the company had discussed a potential bid with its advisers but would likely only proceed if it found a partner for Intel’s manufacturing business.
TSMC, the world’s biggest contract chipmaker, has separately studied controlling some or all of Intel’s chip plants, potentially as part of an investor consortium or other structure, the report said.
Broadcom and TSMC are not working together, and all of the talks so far are preliminary and largely informal, the Journal added.
Intel’s interim executive chairman, Frank Yeary, has been leading the discussions with possible suitors and Trump administration officials, who are concerned about the fate of a company seen as critical to national security, the report said.
Yeary has been telling individuals close to him that he is most focused on maximizing value for Intel shareholders, the report added.
Intel, Broadcom, TSMC and the White House did not immediately respond to Reuters’ requests for comment.
A White House official told Reuters on Friday that President Donald Trump’s administration might not support Intel’s U.S. chip factories being operated by a foreign entity after Bloomberg reported that TSMC was considering taking a controlling stake in Intel’s factories at Trump’s request.
The White House official said the Trump administration supported foreign companies investing and building in the U.S. but was “unlikely” to support a foreign firm operating Intel’s factories.
Bloomberg reported that Trump’s team raised the idea of a deal between the two firms in recent meetings with officials from TSMC, who were receptive, citing a person familiar with the matter.
Intel was among the largest beneficiaries of the U.S. push to onshore critical chip manufacturing led by former President Joe Biden’s administration.
In November, the U.S. Commerce Department said it was finalizing a $7.86 billion government subsidy for Intel.
The company is one of a few chipmakers that design and manufacture semiconductors.
TSMC boasts a market valuation about eight times larger than that of Intel. The Taiwanese company’s customers include AI chip leader Nvidia
and AMD
, which is Intel’s fierce rival in PC and server markets.
Former Intel CEO Pat Gelsinger, who was ousted last year, set sky-high expectations for Intel’s manufacturing and AI capabilities among major clients but fell short, leading to the chipmaker losing or canceling contracts, Reuters reported previously.
Intel’s shares lost about 60% of their value last year as the company’s capital-intensive bid to bolster manufacturing — a strategy championed by Gelsinger — strained its cash flow and ultimately led to it cutting about 15% of its workforce.
doc2016
3日前
geminia 2.0 says, "While specific details about the internal architecture of the NVIDIA mini supercomputer, Project DIGITS, haven't been fully disclosed, it's highly probable that it utilizes stacked die technology. Here's why:
High Performance in a Small Form Factor: Project DIGITS is designed to be a compact, personal AI supercomputer delivering petaflop-level performance. Achieving this level of computing power in such a small device necessitates the use of advanced packaging technologies like stacked dies to maximize chip density and minimize power consumption.
GB10 Grace Blackwell Superchip: The supercomputer is powered by the GB10 Grace Blackwell Superchip, which integrates a Blackwell GPU and a Grace CPU. Both of these components are likely built using stacked die technology to enhance their individual performance and enable efficient communication between them.
Industry Trend: The semiconductor industry is increasingly relying on stacked die technology to create more powerful and efficient chips. This trend is driven by the need to continue Moore's Law and deliver ever-increasing performance. NVIDIA is at the forefront of this trend, and it's logical to assume they would employ this technology in their latest supercomputer.
Memory and Interconnect: High-bandwidth memory (HBM) is often used in conjunction with stacked die technology to provide the necessary memory bandwidth for high-performance computing. Additionally, advanced interconnect technologies are required to enable efficient communication between the stacked dies and other components on the chip.
While we don't have explicit confirmation, the combination of high performance, small form factor, and the use of the GB10 Grace Blackwell Superchip strongly suggests that Project DIGITS leverages stacked die technology."
north40000
4日前
NVDA's attempt to acquire ARM was nixed by DOJ/FTC. A day or two ago, CNBC-TV reported that NVDA dumped its investment in SOUN, and bought shares of other of its partners HIMS and NBLS. I added to positions of both in our portfolio with cash received from mergers involving our ~ 52k shares of CDMO (and fewer shares of RVNC) that completed in the past couple of weeks. CDMO had been #4 in market value in our portfolio, behind JNJ, NVDA, and VCEL.