CoreWeave, founded as a GPU-centric cloud computing provider, has found itself in the right place, at the right time. The massive growth in demand for cloud-based artificial intelligence (AI) capabilities has meant that their specialization in AI and machine learning (ML) workloads has made them a leading competitor in the AI space.
The demand for fast, efficient, high-performance computing has led to the rapid scaling of CoreWeave’s infrastructure, making it one of the primary suppliers of cloud resources for AI applications to a demand-driven market segment.
The AI Data Center Boom
At the end of August, Chirisa Technology Parks announced a $5 billion joint venture with Blue Owl Capital and PowerHouse Data Centers to develop large-scale AI and HPC data centers for CoreWeave and potentially other clients.
Developments are planned across key markets that include New Jersey, Pennsylvania, Texas, Nevada, and Kentucky, with the first phase of this venture being a new data center on Chirisa’s 350-acre data center campus near Richmond, Virginia, a facility which will offer 120 MW of capacity.
Current plans are for the data center space to become available in 2025 and 2026.
This is the first stage of a build-to-suit program which will take advantage of Chirisa’s infrastructure solutions, which have included powered shell and turnkey facilities for a range of customers.
According to Lee Hayes, President and CEO of Chirisa Technology Parks:
"This breakthrough transaction with Blue Owl, CTP and PowerHouse is the first stage of a large-scale partnership focused on meeting the near-term requirements of our hyperscale customer base. Through innovative structuring, broad financial distribution and the transactional expertise of our partners at Blue Owl, this partnership unlocks very substantial growth capital for one of the most important players in today's AI hyperscale landscape."
Blue Owl is an asset management business handling close to $200 billion, with a focus that includes growth tech companies, an area of expertise that reflects the current state of the data center business.
Doug Fleit, CEO and Co-Founder of PowerHouse Data Centers, highlighted the advantages of the partnership driving this joint venture, saying:
"PowerHouse is proud to partner with Blue Owl and CTP. By bringing our proven real estate leadership and cadre of acquisition, development, and power procurement experts to the JV, this partnership will redefine the landscape of precision hyperscale infrastructure."
The Importance of Building for AI Workloads
With the demand for AI-driven computing coming from all sectors, we are seeing its impact on how data centers are designed and provisioned.
The high-performance workloads for AI are changing the industry's approach to many aspects of data center design, with technologies such as liquid cooling and low-latency networking infrastructure seeing significant importance as basic functions.
The AI workload has changed the way data centers approach the physical infrastructure model.
CoreWeave has positioned itself to take advantage of the AI boom by providing an infrastructure that is optimized for this kind of workload, having designed its infrastructure around GPU computing, the critical core of AI operations.
By partnering with Chirisa and its affiliates, CoreWeave is ensuring that its infrastructure is housed in data centers specifically designed to support these high-performance workloads.
Traditional data center designs, which were built to handle more generalized cloud workloads, are being found not to be well-suited to the unique demands of AI without significant changes to power and cooling -- tasks which can be very difficult to be done in a cost-effective manner -- leading to the growth in AI-specific data center development.
The September announcement from the from the White House of a Task Force on U.S. Leadership AI Data Center Infrastructure has a goal of streamlining coordination of policies to advance AI data center development.
The new Task Force is led by the National Economic Council, National Security Council, and the White House Deputy Chief of Staff’s office, along with businesses identified by the government as AI infrastructure leaders, including CoreWeave, in acknowledgement of the company's pivotal role in the industry.
CoreWeave Continues its Expansion
CoreWeave has said that they expect to have 28 data center locations online by the end on 2024, with sites in strategic areas throughout the U.S.
And it’s not simply a matter of finding new colocation partners or using existing facilities. CoreWeave is taking steps to address the important issues of power and cooling throughout their enterprise.
In July of this year, we reported on the strategic partnership that they entered into with Bloom Energy. At that point they had committed to deploying Bloom’s solid oxide proprietary fuel cells to generate power for the CoreWeave high performance data center in Volo, Illinois, which is owned by Chirisa Technology Parks, giving some indication of a potential source of energy for the new, planned, data center developments.
Also, in June of this year, CoreWeave announced their plans to invest $2.2 billion to develop new data centers, powered completely by renewable energy, in Norway, Sweden, and Spain by the end of 2025.
This is a follow-up to their investment of $1.3 billion in a United Kingdom headquarters for their European operations, and in two data centers in the UK.
In the last seven years, CoreWeave has operated a constantly increasing family of data centers in every region of the US and across Europe.
The $5 billion joint venture with Blue Owl Capital and PowerHouse Data Centers, along with the billions in additional development investment, highlights the interest and importance of AI-specific data center development that is currently ongoing.
As discussed above, the growing demands of AI, both from the user side and the hosting side, are driving a reconsideration of what a data center is, how they should be designed, and where they should be developed.
Expansive investments such as we've recounted here are just the tip of the iceberg, and will continue to grow in the data center industry's near term.
CoreWeave and Nokia
And also, technology partnerships. Just yesterday it was announced that Nokia was selected by CoreWeave to provide the networking background behind CoreWeave's hyperscale AI cloud.
CoreWeave will deploy Nokia IP and optical platforms in data centers across the U.S. and Europe as part of what the company characterizes as a "massive wide area network buildout to support high-performance AI infrastructure."
CoreWeave emphasized how the Nokia platforms' ultra-fast, reliable performance at scale supports the lossless, low-latency communications so critical to enabling the AI boom.
Jim Julson, Director of Networking at CoreWeave, said:
“CoreWeave has chosen Nokia hardware to power its backbone and edge platforms to meet the performance, stability and scalability demands that today's and tomorrow’s AI and ML hyper scale clouds require. With the explosion of demand centered around the infrastructure required to meet these demands, Nokia has proven to be a critical partner and we have no doubt that as CoreWeave continues to scale, Nokia will be there to help facilitate a world-class cloud experience for all our customers.”
According to a press release, CoreWeave’s IP backbone is being built with the FP5-based Nokia 7750 Service Router (SR), which provides massive routing scale, 800Gb/s speeds with strong Ethernet VPN (EVPN) support, and Nokia’s high-performance Service Router Operating System (SROS).
Optical transport and data center interconnection across the WAN will leverage the Nokia 1830 Photonic Service Interconnect (PSI) solution. The Nokia Network Services Platform (NSP) will automate network functions and optimize resource allocation.
Vach Kompella, Senior Vice President and General Manager of IP Networks business at Nokia, concluded:
“As the hyperscaler behind some of the biggest AI enterprises and labs in the world, CoreWeave is at the forefront of innovation in unexplored territory. That requires proven reliability, performance at scale without compromise, and the increased efficiency Nokia’s automation solutions bring.
This approach secures lossless and ultra-low latency delivery every single time – regardless of traffic peaks or unexpected events. We are looking forward to deploying Nokia IP and optical platforms in CoreWeave’s backbone as it expands its global network of data centers.”
CoreWeave and the Future of Data Center Design
In the following video from this year's NVIDIA GTC 2024 event, CoreWeave Chief Strategy Officer Brian Venturo discusses the future of data center design and CoreWeave's rapidly expanding presence in the sector.
