The brain becomes the fastest host in the world for Deepseek R1, as NVIDIA GPUS exceeded 57x

Brain systems Today it will be announced that it will host a Dibsic penetration Artificial intelligence model R1 on American serversIncreasing the speed of up to 57 times faster than GPU solutions while maintaining sensitive data within the American border. This step comes amid increasing concerns about rapid artificial intelligence progress in China and data privacy.

The starting start of the artificial intelligence chip will publish a copy of the teacher 70 billion Deepsek-R1 It works on its devices on its own chip, providing 1,600 code in the second-a significant improvement in the traditional GPU applications that have struggled with the latest “thinking” models.

Why do the thinking forms of Deepseek restore the AI’s organization

“Thinking models affect the economy,” James Wang, chief executive of Kerbras, said in an exclusive interview with Venturebeat. “Any knowledge factor must basically do a kind of multi -step cognitive tasks. These logical models will be the tools that enter their workflow.”

This announcement follows a turbulent week in which the appearance of Deepseek led to the appearance of Nvidia’s The biggest loss of market value, approximately $ 600 billionAsking questions about the giant AI’s superpower. The brain solution directly addresses two major concerns that have appeared: the mathematical requirements for advanced artificial intelligence models and data sovereignty.

“If you use Deepseek applications interface“This is very popular at the present time, the data is sent directly to China,” said Wang. “This is a severe warning (making) many American companies and institutions … not to look (IT).”

How the brain technology excel

CEREBRAS achieves the advantage of speed through a new chip structure that maintains the entire artificial intelligence models on one processor of the size of the Woss, which eliminates memory bottlenecks that depend on GPU systems. The company claims that its implementation of DeepSeek-R1 matches or exceeds the performance of Openai’s royal models, while operating it entirely on American soil.

Development is a major shift in the scene of artificial intelligence. DibsicFounded by Liang Winfing, the executive director of the former hedge fund, the industry was shocked by achieving advanced artificial intelligence capabilities by only 1 % of the cost of American competitors. The brain hosting solution now provides American companies a way to take advantage of these developments while maintaining data control.

“It is in fact a beautiful story that American research laboratories gave this gift to the world. Wang said:“ The Chinese have taken it and improving it, but it has restrictions because it works in China, and it has some control problems, and now we return it and turn it on American databases, without controlling, without keeping the data “.

American technology leadership faces new questions as creating artificial intelligence goes globally

The service will be available through Development preview Starting today. Although it will be free at first, the plans are planning to implement API access control tools Because of the strong early demand.

This step comes when American lawmakers are struggling with the effects American trade restrictions Designed to maintain technological advantages on China. The ability of Chinese companies to achieve the capabilities of Amnesty International, however Controls of the export of the chip The invitations have prompted new organizational curricula.

Industrial analysts suggest that this development can accelerate the shift from the GPU’s infrastructure. “NVIDIA is no longer the leader in the performance of reasoning,” Wang pointed out, noting the standards that show superior performance from various specialized artificial intelligence chips. “These are other artificial intelligence chips faster than graphics processing units to operate these recent models.”

The influence extends beyond technical standards. Since artificial intelligence models include increasingly advanced thinking capabilities, their arithmetic demands have risen. Kerbras argues its own elegance with the elegance of these emerging burdens, and perhaps reshaping the competitive scene in the publication of the AI.