Beidian Drives Domestic Computing Expansion with AI

In the era of the digital economy, computational power is nothing short of paramountAs various industries turn toward advanced technologies, it has become clear that the momentum of economic growth is intrinsically tied to the capabilities of computing resourcesMuch like electricity and heating revolutionized industrial landscapes, computational power is emerging as the new driving force behind high-quality development and sustainable growth in the economy.

Amidst this transformative wave, domestic computational capabilities are poised at a remarkable junctureThe rise of large models and artificial intelligence technologies has not only catalyzed an immense demand for computation but has also accelerated the evolution of the domestic computing industryNevertheless, challenges remain; fragmented demand, insufficient communication between supply and demand sides, and an underdeveloped ecosystem hinder the usability of domestic computational power.

So, how can we break the deadlock from the market side? How can we propel domestic computational power forward by encouraging market-oriented approaches? A recent initiative from Beijing Electronic Intelligence Technology Co., Ltd. (referred to as North Electric Intelligence) offers a potential solutionThey have launched the first domestic computational power Proof of Concept (PoC) platform, now open to the public and established within Beijing’s digital economy computational power centerUtilizing North Electric Intelligence's advanced computational iteration verification platform, this initiative provides tailored scene evaluation, compatibility testing, and validation services for domestic computational clusters while fostering AI application development across diverse fields including finance, governance, industry, healthcare, embodied intelligence, and transportation.

The time is ripe for a new vision, one that transitions domestic computational power from merely "usable" to "highly usable." Data points show that the Chinese large model market will reach a scale of 10.5 billion in 2023, projecting to grow to 16.5 billion in 2024, reflecting a staggering year-on-year increase of 57%. By 2028, it is anticipated that the market will scale further to 62.4 billion, driven predominantly by the pressing demand for industrial intelligence transformation.

The integration of artificial intelligence and data elements signifies a critical focus for numerous sectors transitioning and upgrading their operations

Advertisements

The melding of large models with real-world scenarios is no longer a trend; it is necessary for survival in an increasingly competitive environmentThis burgeoning need for computational resources stands as a pivotal opportunity for the advancement of the domestic computational power industry, compelling an ongoing enhancement of capabilities.

In recent years, the rapid development of the domestic computational power industry has led to a surge of home-grown computational chipsWhile this influx has injected vibrancy into the market, it has also faced criticismMany of these chips only achieve a level of competence that falls short of addressing the broader needs of large-scale commercial applicationsAccording to Yang Zhen, CMO and head of strategy and market at North Electric Intelligence, although many domestic chips showcase certain advantages on various metrics, they lack the coherent commercial ecosystems essential for widespread adoption.

One prominent challenge identified includes the deficiency of comprehensive application scenario demonstrations coupled with effective testing mechanismsDue to this gap, users from the power demand side are often left with a vague understanding of domestic computational options available to themTraditional performance testing, primarily focused on theoretical specifications, fails to communicate how these chips perform in varied business contexts, leaving prospective buyers wary of their practical use.

Moreover, as AI applications proliferate, diverse and heterogeneous computational resources have become essential for large-scale training operationsHowever, many domestic data centers face the hurdles associated with multi-chip interoperability, which leads to software incompatibilities, disconnected computational resources, and difficulties in unifying development toolchains.

To address this situation, the North Electric Intelligence strategy emphasizes establishing platforms, creating bridges, and facilitating the flow of demand toward practical applications

Advertisements

Through the domestic computational power PoC platform and the “testing to promote usage” innovation model, the company seeks to connect users with various suppliers of domestic AI chips, enabling them to discover the optimal combinations of chips and software suited for diverse scenariosThis initiative aims to expedite the commercial application of domestic computational technologies by fostering an ecosystem rich in practical applications.

Yang Zhen asserts the presence of numerous large model enterprises and over thirty AI chip companies within ChinaYet, each operates in siloed conditions leading to a fragmented ecosystem that inhibits growthNorth Electric Intelligence’s ambition is to dismantle these silos at every level to facilitate the progression of domestic computational power and establish a holistic commercial framework.

Within the domestic landscape, bridging the gap between computational capabilities and application scenarios is a multifaceted endeavorIt inevitably hinges on building robust foundation layers in infrastructure such as data centers to provide comprehensive testing environments where domestic computational power can thriveThe Beijing digital economy computational power center embodies this vision, boasting not only a strategic geographical position and potent computational scale but also achieving a Power Usage Effectiveness (PUE) rating of 1.146. This facility can offer an incubator environment for both computational power and application scenarios.

Currently, many domestic AI chip companies operate on a smaller scale with limited funding resources, facing high costs tied to chip validation and significant gaps between theoretical applications and practical implementationThus, North Electric Intelligence’s PoC platform aims to demonstrate the actual capabilities of domestic chips across diverse AI scenariosThis showcases the best application environments for the chips and importantly provides AI application developers with performance evaluation services tailored to specific industry needs, actively promoting the use of domestic chips in various sectors.

It has been reported that the Beijing digital economy computational power center has successfully tested deployments for over ten renowned domestic chip manufacturers, including Moore Threads and Cambricon

Advertisements

Additionally, North Electric Intelligence is engaged in validating various industry-specific application scenarios spanning film and media, finance, governance, and manufacturing, thereby setting benchmarks while exploring optimal compatibility among domestic computational power, industry needs, models, and applications.

For example, within the film industry, North Electric Intelligence is implementing an AIGC audiovisual intelligence solution for a cultural model, partnering with over ten leading AI enterprises to develop advanced applications like AI-generated characters and story creation aidsThese efforts aim to facilitate the digital and intelligent transformation of the film and cultural sectors.

Further reflecting their commitment to seamlessly integrating computational performance with AI frameworks, foundational models, and practical applications, North Electric Intelligence has invested considerable resources in optimizing algorithms and enhancing performance across the boardAs a result, their initiatives have reportedly improved domestic chip performance by two to three times and boosted throughput by seven to eight times due to software-level accelerations.

The significance of this dual-channel connection between domestic computational power and application scenarios cannot be overstatedThe launch of the first domestic computational power PoC platform stands as a testament to the capabilities of the Beijing digital economy computational power center, further validating its role in fostering an expansive ecosystem and empowering various verification capabilities for AI heterogeneity.

To summarize, as management guru Ram Charan highlights, structural uncertainties arising from profound changes within the external environment can dramatically reshape industry landscapes and exert substantial impactsThe current climate presents significant challenges for the domestic computational industry in the face of structural uncertainties

Advertisements

Advertisements