The recent upheaval instigated by DeepSeek, a Chinese artificial intelligence startup, has sent shockwaves through Wall Street, igniting an unprecedented debate surrounding the future of AIDubbed the "Sputnik moment" of AI by investors, DeepSeek's large model has cast a long shadow over chip manufacturing giant NVIDIA, prompting discussions centered around what many are now calling a historic "dark hour" for the company.
Following the close of trading on January 27, NVIDIA witnessed a dramatic decline of more than 17 percent in its stock price, shedding over $560 billion in market capitalizationThis collapse dropped the company's valuation below the $3 trillion mark, effectively erasing any gains it had amassed in the year to dateThis rapid plunge marked the largest single-day drop in NVIDIA's market value in its history, eclipsing the previous record set in September 2024 when the stock fell by 9 percent.
The fallout from this dramatic market event did not spare NVIDIA’s co-founder and CEO Jensen Huang, who saw his personal fortune diminish by over $13 billion in a single night
Advertisements
That dramatic loss illustrates how intertwined the fates of tech executives are with the performance of the companies they lead in an industry marked by volatility.
In response to the stock market's tumult on January 28, NVIDIA addressed the sharp fluctuations, asserting that DeepSeek represents a significant advancement in AI technology, showcasing an exemplary case of testing and expansionThe statement emphasized how DeepSeek leveraged widely available models and computational capabilities fully compliant with export regulations to develop new modelsNVIDIA clarified that the inference process requires a substantial number of their GPUs and high-performance networking, reiterating their three expansion laws: the laws of pre-training, post-training, and new rules for scaling during testing.
Meanwhile, as the global market reeled from NVIDIA’s stock performance, DeepSeek made headlines by releasing its multiplatform AI model Janus-Pro, comprising 1 billion and 7 billion parameters
Advertisements
Notably, the Janus-Pro-7B outperformed OpenAI’s DALL-E 3 and Stable Diffusion in GenEval and DPG-Bench benchmark tests — a significant achievement fueling further intrigue in the AI landscape.
On the same day, DeepSeek temporarily restricted new user registrations, citing a "massive malicious attack." This incident caused the application to slip from the number one spot on Apple's download rankingsThe sudden influx of users exceeded expectations, triggering heightened scrutiny and challenges for the startup fresh on the scene.
Many observers now wonder if DeepSeek has managed to puncture the much-lauded AI bubbleWhile existing users can still log into the platform, some have reported discrepancies in the model’s responses, particularly in niche fields requiring specialized knowledgeComplaints highlighted instances of the model generating references to non-existent literature, a situation potentially rooted in the limitations of the training corpus and datasets employed.
Regardless, the emergence of DeepSeek has sparked a profound discourse on AI's developmental trajectory, forcing market participants to reassess the direction of AI evolution
Advertisements
The intense reaction from Wall Street underscores just how inflated the AI bubble had become, especially considering that prior to DeepSeek's introduction, companies like OpenAI and Anthropic had received astronomical levels of funding and secured valuations worth hundreds of billions of dollars.
If DeepSeek's more affordable model proves viable, it could entirely upend the current investment logic governing the AI sectorThe implications would be seismic, likely rendering the vast investments of major American tech firms superficial and calling into question the valuations that have soared in recent monthsThe wild fluctuations in NVIDIA's stock price further indicate that the valuations within the AI chip industry may need substantial reevaluation.
Historically, NVIDIA has positioned itself as the primary beneficiary of the generative AI wave, with its GPU chips forming the backbone of countless data centers across the globe
- US Stocks Continue to Rise in January
- Nasdaq Surges Over 2%
- US-Canada Trade Spat: Will Oil Prices Surge?
- Bank of Japan Resumes Interest Rate Hikes
- FOF Asset Allocation: Strategies for Steady Progress
As tech companies worldwide continue to tussle for NVIDIA's leading-edge AI chips, these products are often prohibitively expensive, sometimes costing tens of thousands of dollars each.
Fueled by fervent investor enthusiasm, NVIDIA surged past a stock market valuation of $3 trillion at a record-breaking paceThe company’s stock price doubled within the last year, and prior to the recent plunge, Jensen Huang’s net worth peaked at $130 billion, securing his spot on the Forbes billionaire list as the ninth wealthiest individual.
Whether NVIDIA can retain this lofty valuation is a question rapidly gaining tractionHuang has consistently conveyed confidence in the strong demand for GPU chips, stating that as long as the movement toward building global data centers continues, dependence on NVIDIA’s technology remains intactHowever, if a competitor emerges proving that it is possible to train robust AI models without expenditure on such costly chips, the validity of NVIDIA’s $3 trillion valuation may collapse.
In Huang's view, the global data center market was valued at $250 billion a year prior and is projected to grow at 20 to 25 percent annually, which has historically supported NVIDIA's valuation narrative
Meanwhile, UBS estimates that investments in AI among major U.Stech firms reached $224 billion last year and could escalate to $280 billion by 2025, with much of this capital directed toward constructing data center infrastructures where GPUs play a central role.
Gartner analyst Sheng Linghai has noted that while previously there was little question that computational power dictated AI advancement, perceptions are shiftingHe suggests that DeepSeek’s achievements may imply that AI's capabilities are not entirely dependent on computational resourcesIf substantiated, this finding could significantly narrow the market space for expansive AI training accelerators, thus shrinking NVIDIA’s future prospects.
While no major U.Stech entity has formally declared a pivot away from high-commitment computational resources, Meta has announced plans to increase capital expenditures on AI projects by around 50% to $65 billion this year
Nevertheless, skepticism surrounding the sustainability of this "burning money" strategy is beginning to surface in capital markets.
Raymond James semiconductor analyst Srini Pajjuri observed in a recent report that while DeepSeek may not have access to the same levels of computing power as U.Stech giants, its ability to create a highly competitive model may stimulate a renewed urgency among these larger enterprisesHowever, he believes these firms may leverage their substantial GPU caches to distinguish themselves from lower-cost entries like DeepSeek.
Citigroup analysts indicated that the emergence of DeepSeek has elevated the conversation surrounding computational costs to a "critical topic." They further asserted that U.Stech firms are unlikely to abandon their pursuit of advanced chips, especially given the federal commitment of $500 billion towards the "Stargate" AI infrastructure project underlining the ongoing demand for state-of-the-art chips.
Optimistic analysts suggest that although DeepSeek's rise may exert short-term pressure on NVIDIA’s stock, it could eventually reduce the barriers and costs surrounding AI large models, potentially increasing the demand for computational resources in the longer term, resulting in favorable conditions for AI chip investments.