What is Neocloud?

Neocloud

In tech terms, a neocloud is a new breed of cloud infrastructure purpose-built for AI and high-performance computing (HPC).

Unlike traditional hyperscale cloud providers (like AWS or Azure), neoclouds focus on delivering raw GPU power, low-latency performance, and specialised environments for compute-intensive workloads.

🧠 Key Features of Neoclouds

  • GPU-as-a-Service (GPUaaS): Optimised for training and running large AI models.
  • AI-native architecture: Designed specifically for machine learning, deep learning, and real-time inference.
  • Edge-ready: Supports distributed deployments closer to users for faster response times.
  • Transparent pricing: Often more cost-efficient than hyperscalers for AI workloads.
  • Bare-metal access: Minimal virtualisation for maximum performance.

🏗️ How They Differ from Traditional Clouds

FeatureNeocloudsHyperscale Clouds
FocusAI & HPC workloadsGeneral-purpose services
HardwareGPU-centric, high-density clustersMixed CPU/GPU, broad service range
FlexibilityAgile, workload-specificBroad but less specialised
LatencyUltra-low, edge-optimizedHigher, centralized infrastructure
PricingUsage-based, transparentOften complex, with hidden costs

🚀 Who Uses Neoclouds?

  • AI startups building chatbots, LLMs, or recommendation engines
  • Research labs running simulations or genomics
  • Media studios doing real-time rendering or VFX
  • Enterprises deploying private AI models or edge computing

Think of neoclouds as specialist GPU clouds—like a high-performance race car compared to a family SUV.

Both get you places, but one’s built for speed, precision, and specialised terrain.

What is China’s equivalent to Nvidia?

AI microchips

Chinese firms are reportedly intensifying their efforts to develop a competitive alternative to Nvidia’s AI chips, as part of Beijing’s ongoing initiative to reduce its reliance on U.S. technology.

China faces several challenges that are impeding its technological progress, including U.S. export restrictions that limit domestic semiconductor production. The lack of technical expertise is also reported to be a problem.

Analysts have identified companies including Huawei as the principal competitors to Nvidia in China

China’s counterparts to Nvidia, such as Huawei, Alibaba, and Baidu, are actively developing AI chips to compete in the same market. Huawei’s HiSilicon division is known for its Ascend series of data centre processors.

Huawei’s HiSilicon division is known for its Ascend series of data centre processors, and Alibaba’s T-Head has produced the Hanguang 800 AI inference chip. Other significant players include Biren Technology and Cambricon Technologies.

Alibaba’s T-Head has developed the Hanguang 800 AI inference chip. Other significant players include Biren Technology and Cambricon Technologies.

These Chinese firms are intensifying their efforts to create alternatives to Nvidia’s AI-powering chips. This is a big part of Beijing’s broader initiative to reduce its reliance on U.S. technology.

Nvidia’s surge in growth is attributed to the demand from major cloud computing companies for its server products, which incorporate graphics processing units, or GPUs.

These GPUs are crucial for entities like OpenAI, the creator of ChatGPT, which requires substantial computational power to train extensive AI models on large datasets.

AI models are crucial for chatbots and other AI applications

Since 2022, the U.S. has limited the export of Nvidia’s top-tier chips to China, with further restrictions imposed last year.

The U.S. sanctions and Nvidia’s market dominance pose significant obstacles to China’s ambitions, particularly in the short term, according to analysts. The U.S. has curbed the export of Nvidia’s most sophisticated chips to China since 2022, with increased restrictions implemented last year.

China’s GPU designers rely on external manufacturers for chip production. Traditionally, this role was filled by Taiwan Semiconductor Manufacturing Co. (TSMC). However, due to U.S. restrictions, many Chinese firms are now unable to procure chips from TSMC.

As a result, they have shifted to using SMIC, China’s largest chipmaker, which is technologically several generations behind TSMC. This gap is partly due to Washington’s limitations on SMIC’s access to essential machinery from the Dutch company ASML, necessary for producing the most advanced chips.

Huawei is driving the development of more sophisticated chips for its smartphones and AI, which occupies a significant portion of SMIC’s capacity.

Nvidia has achieved success not only through its advanced semiconductors but also via its CUDA software platform. The system enables developers to build applications for Nvidia’s hardware. This has fostered an ecosystem around Nvidia’s designs, which will be challenging for competitors to emulate.

Huawei leading the pack for China

Huawei is at the forefront as a leading force in China for its Ascend series of data centre processors. The current generation, named Ascend 910B, is soon to be succeeded by the Ascend 910C. This new chip may come to rival Nvidia’s H100.

Nvidia announces new AI chips

AI

Nvidia has revealed its latest generation of AI chips, coming just months after the release of its preceding model.

This rapid succession underscores the intense competition within the AI chip market and Nvidia’s relentless effort to maintain its leading position.

CEO Jensen Huang has now committed to unveiling new AI chip technology annually, accelerating the company’s prior biannual pace. The latest AI chip architecture, named ‘Rubin,’ is set to follow the ‘Blackwell’ model announced in March 2024, which is currently in production and anticipated to be delivered to customers the latter part of 2024.

Huang’s unveiling of the Rubin has seemingly hastened Nvidia’s already rapid AI chip development.

Nvidia has committed to launching new AI chip designs annually, a cadence Huang reportedly referred to as a ‘one-year rhythm‘ during his Sunday 2nd June 2024 announcement. Previously, the company was committed to updating its chips every two years. But such is the speed and fierce competition of AI development, that original decision has become quickly out-dated.

The swift transition from Blackwell to Rubin, taking less than three months, highlights the intense competition in the AI chip market and Nvidia’s race to maintain its leading position.

AMD and Intel are two major competitors playing catch-up in the AI race.

Nvidia one year share chart

Nvidia one year share chart

Wall Street is in love with Nvidia as results beat estimates – but can the stock maintain these meteoric gains?

AI GPU

Nvidia’s shares surpassed $1,000 for the first time during extended trading on Wednesday 22nd May 2024, following the chip manufacturers report of fiscal first-quarter (Q1) earnings that exceeded analysts’ expectations.

Investors have been using Nvidia’s performance as a barometer for the AI industry’s growth, which has captivated the market over the past year. The robust results indicate that the demand for Nvidia’s AI chips continues to be strong. However, there may be an argument that it is time to take some profits from these massive gains. Can it continue its meteoric climb?

It was also announced that revenues from the upcoming next-generation AI chip, ‘Blackwell‘, are expected later in the year.

In extended trading, the stock increased by around 7%. Additionally, Nvidia announced a 10-for-1 stock split. Given the post-market activity, the shares are on track to reach a new high on Thursday 23rd May 2024.

Nvidia one year share price

Nvidia one year share price

Earnings Per Share: $6.12 vs. $5.98 – (Nvidia financial reports)

Revenue: $26.04 billion vs. $24.65 billion – (Nvidia financial reports)

Nvidia anticipates sales of $28 billion for the current quarter, surpassing Wall Street’s expectations of $26.61 billion sales, as reported – (Nvidia financial reports)

The company declared a net income of $14.88 billion, or $5.98 per share, a significant increase from $2.04 billion, or 82 cents per share, in the same period last year. (Nvidia financial reports)

Over the past year, Nvidia’s sales have surged, driven by purchases from tech giants like Google, Microsoft, Meta, Amazon, and OpenAI, which have invested billions in Nvidia’s GPUs. These high-end, expensive chips are essential for the development and deployment of artificial intelligence (AI) applications.

Nvidia’s primary business segment, data center sales, encompasses AI chips and other necessary components for operating large AI servers.

The revenue for Nvidia’s data centre sector soared over 400% compared to the previous year. This growth was attributed to the delivery of the company’s ‘Hopper’ graphics processors (GPU’s), including the H100 GPU.

It was also reported that Meta’s Lama 3, their newest large language model utilizing 24,000 H100 GPUs, as a notable income stream this quarter.

Nvidia plan to enhance AI induced success

AI GPU

Nvidia have announced a new generation of artificial intelligence chips and software for running AI models. It’s called: The Blackwell B200 GPU

Blackwell B200 GPU

The Blackwell B200 is the successor to Nvidia’s Hopper H100 and H200 GPUs.

It represents a massive generational leap in computational power.

AI Performance: The B200 GPU delivers 4 times the AI training performance and 30 times the inference performance compared to its predecessor.

Transistor Count: It packs an impressive 208 billion transistors, more than doubling the transistor count of the existing H100.

Memory: The B200 features 192GB of HBM3e memory with an impressive bandwidth of 8 TB/s.

Architecture: The Blackwell architecture takes over from H100/H200.

*Dual-Die Configuration: The B200 is not a single GPU in the traditional sense. Instead, it consists of two tightly coupled die, functioning as one unified CUDA GPU. These chips are linked via a 10 TB/s NV-HBI connection to ensure coherent operation.

*Dual-die packaging technology is used to pack two integrated circuit chips in one single package module. It doubles functionality levels.

Process Node: The B200 utilizes TSMC’s 4NP process node, a refined version of the 4N process used by Hopper H100 and Ada Lovelace architecture GPUs.

The Blackwell B200 is designed for data centres and AI workloads but will likely be available to expect consumer in the future, although these may differ significantly from the data centre model.

Grace Blackwell GB200 Superchip:

Nvidia’s GB200 Grace Blackwell Superchip, with two B200 graphics processors and one Arm-based central processor

This superchip pairs the Grace CPU architecture with the updated Blackwell GPU.

It’s another addition to Nvidia’s lineup, combining CPU and GPU power for advanced computing tasks.

Nvidia continues to push the boundaries of accelerated computing, and these new GPUs promise remarkable performance improvements for AI and other workloads.

Onwards and upwards for Nvidia and the advancement of AI.