In recent discussions surrounding the technological landscape in China, a notable shift has occurred in the intelligent computing industry. While there have been reports suggesting that DeepSeek has caused a halt in the development of various intelligent computing centers, not all industry insiders are in agreement with this narrative. In fact, following the Chinese New Year, several enterprises, particularly publicly listed companies, have begun to integrate DeepSeek into their operations. The coming months are poised to serve as a crucial period for application validation.

Experts in the industry anticipate several key trends emerging in the domestic intelligent computing market this year:

• The large-scale training market is expected to see a decrease in major projects. Decision-making regarding large intelligent computing center projects will likely be delayed by six months to a year, as companies adopt a wait-and-see approach.
• On the inference side of the market, accessible and affordable large models are expected to trigger significant demand from ordinary enterprises. This demand will stimulate growth in cloud major players' inference services, with firms likely to deploy between one to twenty private servers.
• There is a noticeable shift in computation demand from centralization to fragmentation, yet the overall market growth for intelligent computing remains optimistic, suggesting that the establishment of computing centers will continue.
• The requirements for computational power in inference applications are decreasing, which may favor domestic computational capacity. This could lead to the activation of previously underutilized domestic intelligent computing centers.
• The urgency surrounding the adoption of liquid cooling technologies in computing centers may also diminish.
• Notably, Nvidia's stock prices have become intricately linked to the anticipated release of GPT-5, amidst continued robust demand in the domestic market.

As large-scale contracts for intelligent computing are expected to diminish, a veteran server industry professional has expressed skepticism regarding investment in expansive intelligent computing infrastructures. While major internet corporations are not likely to concede the large model marketplace to DeepSeek, it appears that most enterprises are pivoting their focus towards application development. This pivot is contingent upon the extent of secondary development; merely fine-tuning large models will not necessitate extensive computational resources.

DeepSeek has challenged previous assumptions, substantially reducing computational resource consumption and presenting the industry with a new perspective. One insider remarked, "This is undoubtedly a positive development. The idea of increasing computational demands to an unrealistic level, essentially asserting that the entire nation’s electricity should be diverted to it, is simply impractical. This dynamic was never intended for sustainable industrial growth."

There is a growing consensus within the industry indicating that the rush to establish massive clusters of 100,000 or even 500,000 cards, greatly influenced by OpenAI, has lessened in urgency.

Furthermore, investment decisions concerning previously considered super-scale intelligent computing centers, involving billions of dollars, are now being reevaluated. According to one industry veteran, "For substantial intelligent computing projects, we are now entering a six-month to one-year observation phase." Many enterprises focusing on large models are gearing up for a crucial stage of application validation with DeepSeek.

However, the existence of large contracts should not be overlooked. As one representative from a major model enterprise articulated, "DeepSeek is not the end of the road; competition in multi-modal large models and world models is just beginning. These training demands will still begin at tens of thousands of cards."

Transitioning to the inference market, it’s observed that major public cloud providers are eager for growth. According to an industry analyst, "Looking back at the past year, while large models have achieved deployment, we have yet to see explosive growth in inference and industry applications." The affordability offered by DeepSeek could catalyze various industries to accelerate their intelligent applications into a phase of explosive development.

The inference market is bifurcating into public and private cloud domains. Public cloud services predominantly belong to major tech players, and the urgency among them has intensified especially after the launch of ChatGPT, which stirred competition amongst service providers. Notably, upon the release of DeepSeek R1's open-source model, prominent companies like Microsoft and Amazon AWS were quick to announce their offerings, perhaps even preceding their Chinese counterparts, signaling their anticipation for a booming inference market.

Microsoft's CEO Satya Nadella commented on DeepSeek R1's release, stating, "Jevons paradox is being validated again! As AI becomes more efficient and accessible, its usage will skyrocket, turning into an item we cannot get enough of."

An industry analyst, Ben Thompson, has noted that Microsoft's waning partnership with OpenAI largely stems from financial concerns regarding the commitment to inference services versus the substantial investment required in modern data centers. There exists the risk that models may become outdated before they can be effectively commercialized.

As of now, numerous domestic public cloud enterprises and operators have begun integrating DeepSeek into their platforms, provoking a surge of new users eager to experiment, leading to operational pressures on certain cloud platforms.

Moreover, speculation exists that DeepSeek may evolve to offer similar services. Years ago, MongoDB transitioned from an open-source database to a SaaS model, setting a precedent. A seasoned cloud computing practitioner remarked, "Major companies have historically leveraged open-source software for profit; this trend could certainly replicate itself in the domestic market."

Competition is now focused on the private deployment market, with industry experts anticipating a surge. Following the Chinese New Year, several individuals involved in intelligent computing reported that all publicly listed companies are integrating DeepSeek into their operations. These enterprises are exploring localized deployments, tailoring applications to their specific needs in areas like AI-generated personas, games, programming, and sales.

One notable analogy made by an industry expert likened the current shift to the transition from mainframes to PCs, a period when IBM introduced personal computers, democratizing computing access. Many enterprises are now looking to build "small intelligent computing centers," deploying between one to ten servers or possibly up to twenty servers, each with several hundred cards. According to experts, this has become a focal strategic direction presented by IT companies to their clients.

The ramifications of private deployment models extend to sensitive industries; for example, legal firms may hesitate to place client contracts in public cloud environments due to confidentiality risks. Therefore, many opportunities for private deployments are expected to materialize.

Nonetheless, as industry players utilize shared open-source frameworks and standardized hardware, differentiation among technical service providers will intensify, with key focuses on enhancing inference speed and cost-effectiveness. The integration of software and hardware will be pivotal. Collaborative efforts among chip manufacturers, server companies, solution vendors, and private cloud service providers will become increasingly active.

Market dynamics indicate a cautious optimism regarding the intelligent computing sector. If the decline in significant contracts occurs alongside an increase in fragmented orders from private deployments, the overall trajectory of China's intelligent computing market remains uncertain.

However, many maintain a tempered optimism. It’s projected that if smaller enterprises invest in systems priced modestly in the tens of thousands of dollars, with larger entities investing several million, this cumulative growth could reach substantial heights in the foreseeable future. A perennial cloud expert remarked, "DeepSeek is akin to the moment when James Watt revolutionized the steam engine, achieving stable and efficient power output that permeated across various industries."

The advent of large models can be likened to the steam engine's impact after refinement, enabling entry into diverse sectors.

As the market continues to evolve, the importance of insightful application enterprises remains crucial for establishing viable operational models.

As the conversation shifts toward Nvidia, its stock volatility coincides with expectations regarding GPT-5. Upon DeepSeek's R1 release, Nvidia faced a significant stock drop, losing approximately 16.97% of its market value, translating to an evaporation of around $590 billion.

Industry sentiments regarding Nvidia remain contentious, as many view its past performance as unsustainable. The company's high gross margin of 90% raises concerns over long-term viability. Even though major competitors like Microsoft, Google, Meta, and xAI do not plan to reduce capital expenditures, these firms are now exploring the development of in-house chips.

One analyst noted that the quicker rebound of Broadcom’s stock compared to Nvidia's arises from its focus on ASIC chipset production for major clients. These specialized chips offer notable advantages in performance and costs, which may reduce reliance on Nvidia in the future.

It's also understood that markets don't favor monopolistic dominance indefinitely; as illustrated by historical shifts in the CPU market from Intel's stronghold to AMD's resurgence. While Nvidia's dominance is currently pronounced, sustained innovation and market responsiveness will determine its future success.

With the integration of DeepSeek into the CUDA ecosystem, Nvidia's strategy appears to pivot towards more diverse solutions that meet emerging needs. Recent presentations at CES highlighted lower-end products aimed at capturing the inference market, potentially suggesting a strategic readjustment.

Providing a forecast on Nvidia's stock performance hinges primarily on the impact of the GPT-5 release. Market observers claim that a lukewarm response to this model would likely result in significant declines, while a substantial performance would stabilize its financial outlook.

Nevertheless, as competition intensifies among Silicon Valley giants, the inference market will present formidable challenges to Nvidia. Domestic chip manufacturers now have an opportunity to capitalize on these developments, particularly as localized models begin to adapt more efficiently to platforms like DeepSeek.

To seize this momentum, the synergy between domestic chips and innovative software applications remains fundamentally crucial.

Looking ahead, the landscape of intelligent computing is indeed set for intense competition. With the release of DeepSeek inspiring a wave of innovation among small to mid-sized enterprises eager to develop intelligent applications, emergent solutions could create breakout products across various sectors, potentially driving substantial consumption in the inference market.

However, industry experts caution that the hype surrounding AI has escalated artificially high, rendering it a commodity positioned for harvesting profits. While DeepSeek's introduction responds to demand for cost-effective solutions, it remains within a nascent stage where skilled professionals in both AI technology and business are scarce across industries. Without the emergence of outstanding AI applications in the short term, initial enthusiasm may be tempered.

The ever-evolving landscape leads experts to question whether subsequent versions of DeepSeek can match the evolutionary pace of competing international models. Elon Musk acknowledged the impressive capabilities of DeepSeek while positioning future advancements from companies like xAI as imminent possibilities. This highlights that competition will only accelerate as these enterprises leverage DeepSeek technologies, combined with cutting-edge computational support.

As we move forward, the intelligent computing sector in China appears poised for vibrant growth spurred by DeepSeek's innovations, promising a thriving ecosystem where creativity and technological advancement converge. A practitioner in the sector observed how accessible these developments have become, allowing even young students to utilize DeepSeek for creative endeavors, such as programming a simple game. This accessibility heralds a future where anyone can harness computing power for innovation, suggesting that real change is simply a matter of time.