New York
CNN
 — 

Imagine a language tutor who is always on, available anytime to teach new vocabulary or check up on a student’s progress.

On Thursday, Nvidia launched a language learning platform using artificial intelligence that promises to do just that for American Sign Language learners, in partnership with the American Society for Deaf Children and creative agency Hello Monday.

The platform, called Signs, features a 3-D avatar to demonstrate signs. Users keep their video cameras on while interacting with the platform and an AI tool provides feedback as they practice the signs. At launch, the platform features 100 distinct signs, but Nvidia hopes to grow that to 1,000.

Signs represents just one of the many ways AI is helping to advance work on assistive technologies, or tools designed to help people who are disabled or elderly, or their caretakers. Meta, Google and OpenAI have, for example, used AI to improve features for blind or low-vision users, and Apple introduced AI-enabled eye tracking to help physically disabled users navigate their iPhones. Blind users say those advancements are already making it easier for them to navigate life and work.

American Sign Language is the third most prevalent language in the United States, behind English and Spanish, according to the groups behind Signs.

The ASL-learning platform is also a reminder that Nvidia has been trying to branch out into more than just the hardware behind AI. Nvidia has become a major supplier to the AI industry by building the chips that most companies use to run the technology, in addition to its own AI models and software platforms. The company’s stock has soared more than 100% in the past year as AI companies who proclaim the technology’s future promise buy up vast amounts of Nvidia’s chips, bringing it to a more than $3.4 trillion valuation.

Michael Boone, Nvidia’s manager for trustworthy AI product, said the company is committed to building AI products not just for corporate customers, but to also foster practical applications for the technology. “It’s important for us to produce efforts like Signs, because we want to enable not just one company or a set of companies, but we want to enable the ecosystem,” Boone said in an interview with CNN.

And, ultimately, more people using AI in any form is good for Nvidia’s core chip-making business. Some investors have in recent months raised concerns about whether tech giants have been overspending on AI infrastructure, including chips, and questioned how long it might take to earn a return on that investment.

Signs is free to use, and it will allow ASL speakers to contribute videos of signs that are not already available on the platform to grow its vocabulary. That data could also enable Nvidia to develop new ASL-related products down the road — for example, to improve sign recognition in video-conferencing software or gesture control in cars. The company says it will also make the data repository publicly available for other developers.

And for future iterations of Signs, Nvidia said the team behind the platform is exploring how to include non-manual signals that are crucial to ASL, such as facial expressions and head movements, as well as slang and regional variations in the language.

“Most deaf children are born to hearing parents. Giving family members accessible tools like Signs to start learning ASL early enables them to open an effective communication channel with children as young as six to eight months old,” Cheri Dowling, executive director of the American Society for Deaf Children, said in a statement about the new project.

On Jan. 20, Chinese artificial intelligence start-up DeepSeek released its first-generation reasoning models. In the release, the company made some astonishing claims.
First, DeepSeek said its DeepSeek-R1 model achieves performance comparable with OpenAI-o1, widely considered the best-performing model available across most domains. Considering the Chinese company is working with significantly worse hardware than OpenAI and other American companies, that’s certainly remarkable.
Even more impressive is that the company claims to have achieved these results at an incredibly low cost. R1 was built on DeepSeek’s V3 large language model, released in December. The company estimates the compute cost for training V3 came to just $5.6 million. To put that in perspective, OpenAI’s GPT-4 cost $100 million to train.
DeepSeek achieved similar performance at a fraction of the cost. And since it’s completely open source, allowing anyone to copy its techniques, it will have lasting implications on the entire industry.
Two companies, in particular, are in a great position to benefit from DeepSeek’s innovations.

Image source: Getty Images.

The long-term impact of DeepSeek-V3 and R1
DeepSeek focused on maximizing the efficiency of its limited hardware capabilities. Because of AI chip export restrictions, Nvidia isn’t able to sell its most powerful H100 GPUs in China. Instead, it sells H800 GPUs, which are specifically designed to comply with U.S. regulations. The H800 reduces the chip-to-chip transfer rate, reducing the speed at which it’s possible to train large AI models.
Due to such limitations, DeepSeek developed processes that enable it to reduce the amount of data it needs to transfer throughout the system at any given time. For example, its “mixture of experts,” or DeepSeekMoE, introduced last year, made it so it only had to activate part of the model to respond to queries.
DeepSeek isn’t the only company using this method, but its novel approach also made its training more efficient. Most methods involve more training overhead in order to reduce the cost of inference later on.
The start-up also developed methods to reduce the amount of memory required for AI inference by compressing important data before storing and transmitting it. It brought new approaches to load balancing, which is how processes are distributed across its network of GPUs.
The result of these and other breakthroughs isn’t just an AI model that’s faster to train and costs less. The longer-term impact of DeepSeek’s innovations are that it’s cheaper to run, and it can run on less-capable hardware. In other words, AI inference just got a lot more accessible.
In a world with the potential to run AI systems on hardware that fits in your pocket and for a tiny fraction of a penny, there are two very big winners: Apple (AAPL -0.67%) and Meta Platforms (META 0.32%). Here’s why.

Making on-device AI a reality
When Apple started developing its artificial intelligence features for the iPhone and other devices, it put data privacy at the forefront of its efforts. Apple Intelligence is designed to run as much as possible on the iPhone. When it has to make a call to the cloud, it takes every step it can to encrypt user data in the process.
There’s a reason the new AI features Apple introduced last year are only available on iPhones released in the last 15 months. Since Apple is trying to keep everything on the device, it needs enough processing power and memory to run its AI. The newest iPhone chip, the A18 Pro, boosted the memory bandwidth to support faster AI processing.
Apple could adopt many of DeepSeek’s methods to make the iPhone more capable of handling AI inference. That opens the door for features like a more conversational and context-aware Siri, faster translation with no internet connection needed, smart camera features, and better productivity tools. More advanced AI features could boost Apple’s iPhone sales and services revenue.
Apple stock currently trades for a relatively high multiple of 32.5 times forward earnings. But with its massive cash flow, which it uses to buy back shares, and improving profitability from services revenue, it can justify that high multiple, especially considering the consistency Apple has exhibited in recent years. The potential boost from major improvements to on-device AI could be a catalyst for growth over the next few years.

Scaling AI to 3 billion people
Meta’s AI spending is growing fast as it works to scale its capabilities and expand AI features to more parts of its business. Capital expenditures grew about 40% in 2024, and management said it expects a 60% increase in 2025. Those AI investments have paid off well for Meta, resulting in stronger engagement, better advertising tools, and new features like Meta AI, which have the potential to become meaningful sources of revenue down the road.
One important decision Meta made when it came to AI was to open-source its AI model Llama. One of the impetuses behind that decision was to help make the model more efficient. In fact, DeepSeek used Llama as the foundation for developing R1, so this is exactly what Meta hoped for.
Reducing the cost of AI inference could unlock huge profits for Meta. It’s a problem Meta’s been working on for a long time. “A lot of the stuff is expensive, right, to kind of generate an image or a video or a chat interaction,” Zuckerberg said during an earnings call in Feb. 2023. “So one of the big interesting challenges here also is going to be how do we scale this and make this work more efficient so that way, we can bring it to a much larger user base.”
DeepSeek’s answering that challenge and giving Meta the tools it needs to scale AI to its 3 billion users. While Meta might not slow down its spending on AI anytime soon, it’s now capable of making a lot more money off the spending it’s committed to.
Meta stock has zoomed higher on the DeepSeek news, reaching a new all-time high. Still, shares trade for 26.8 times forward earnings estimates as of this writing. Meta’s also a cash cow, using excess free cash flow to buy back shares and support strong earnings-per-share growth. If it can make AI more profitable, it stands to see earnings climb substantially over the next few years, making it well worth the price.

No one thought the path to artificial general intelligence (AGI) would be smooth for investors, but the emergence of DeepSeek has clearly thrown a plot twist into the AI narrative.
Nvidia (NVDA -3.67%) and other AI stocks plunged on Monday, Jan. 27, as investors responded to the threat from DeepSeek, the Chinese AI chatbot that rivals top models like ChatGPT for a fraction of the cost.
Nvidia lost 17% in one session, wiping out $600 billion in market value, the biggest one-day loss for a single stock in market history. Since then, Nvidia has recouped some of those losses, a sign investors may believe the sell-off may have been an overreaction.
Nonetheless, AI stocks remain significantly lower, and Nvidia itself tipped its hat to the Chinese start-up, with a spokesperson calling it “an excellent AI advancement.”
DeepSeek also seems to be gaining credibility, as Microsoft, which is believed to be OpenAI’s biggest investor, has already added the model to its Azure cloud infrastructure service.
So how big of a threat is DeepSeek to the AI ecosystem? To answer that question, let’s outline a few facts about DeepSeek first.

Image source: Nvidia.

What we know about DeepSeek
DeepSeek is a Chinese AI start-up founded by hedge fund chief Liang Wenfeng in May 2023. Unlike OpenAI’s ChatGPT or Alphabet’s Gemini, DeepSeek uses an open-source large language model, meaning developers can update it and adapt it to their own needs.
DeepSeek is significant because its R1 model rivals OpenAI’s o1 in categories like math, code, and reasoning tasks, and it purportedly does that with less advanced chips and at a much lower cost.
According to one estimate, it costs OpenAI’s o1 model $60 to generate a million tokens of output, while DeepSeek’s R1 can deliver the same quantity for just $2.19.
DeepSeek has impressed industry insiders with a 22-page research paper explaining how its model works, but the company has also been accused by OpenAI of using a method called distillation to build its models, a cost-efficient way of training an AI model using larger, more adept ones. Doing so constitutes a violation of OpenAI’s terms of service.
Distillation is commonly used in AI, but if that accusation is true, it would seem to undermine a lot of DeepSeek’s credibility, making it seem like the Chinese start-up plagiarized at least part of its model.
There’s also a debate over how much DeepSeek actually paid for its infrastructure, as it said it cost just $5.6 million to train its V3 model. The V3 was built on Nvidia H800s, which were made to get around U.S. export rules and perform similarly to H100s, Nvidia’s GPUs that have been widely used to build AI infrastructure and models in the U.S.
Analysts have cast doubt on the $5.6 million figure, and that doesn’t seem to include essential costs like research, architecture, or data, making it difficult to do a direct comparison with U.S-based AI models that have required billions of dollars in investments.
What it means for Nvidia
It’s too early to know what the implications of DeepSeek are for Nvidia and the broader AI sector, and there’s still a lot of uncertainty around what exactly DeepSeek has achieved.
The company appears to have made genuine gains in efficiency, but those seem less impressive if its model was built in part by borrowing from OpenAI. The true cost of the model also isn’t fully clear.
The DeepSeek-R1 launch was called a “Sputnik moment” by Silicon Valley honcho Marc Andreessen and others, and the geopolitical implications of the new chatbot could be just as meaningful as the technological ones.
The U.S. could respond by intensifying the tech cold war with China, tightening export rules further and taking other measures. Additionally, allowing DeepSeek on U.S. smartphones while banning TikTok seems incongruous, and U.S. corporations and governments are likely to be skeptical of handing their data over to a Chinese start-up.
For Nvidia investors, it’s also worth remembering that this is just one episode in a years-long technology evolution, and is probably not as meaningful as a $600 billion one-day sell-off makes it seem. Even if DeepSeek shifts the entire industry to a more efficient open-source architecture, that could be a positive for Nvidia over the long run. According to Jevons paradox, lowering the price to run AI models could increase demand, leading to an increase in total consumption, which would drive more purchases of AI chips from Nvidia, though likely at a lower cost.
It’s also meaningful that DeepSeek was built on Nvidia chips. No one’s challenging its supremacy there.

Why I’m not worried about Nvidia
Investors shouldn’t miss the forest for the trees here. You should remember that a competent AI chatbot isn’t the goal here. The ultimate goal is artificial general intelligence, including applications like autonomous vehicles and robotics, and it’s unclear if DeepSeek dramatically changes the calculus around that.
Those technologies are powerful and valuable enough that the race toward AGI will continue, and the tech giants competing in it will continue to pour billions into the infrastructure necessary to build it. Efficiency is important, but technological leadership is the real prize here.
Nvidia stock was already dealt a setback by DeepSeek, and that could be true of Nvidia’s business as well, but the company has proven itself to be nimble before.
It’s evolved its technology to go from primarily serving video games to cryptocurrency mining, AI, autonomous vehicles, 3D rendering, and more. CEO Jensen Huang is rightly regarded as a visionary in the industry, and it continues to rapidly innovate with its new Rubin platform in development.
The AI frontier will continue to evolve, and Nvidia will adapt to market conditions as needed. Whatever the impact of DeepSeek, the race to AGI isn’t going away, and neither is Nvidia.