THE APEX TIMES
Firmus and NVIDIA plan 170,000-GPU AI data center in Indonesia, with deployments extending through 2028
The project, backed by NVIDIA’s data center platform, is set to roll out Vera-Rubin, Grace-Blackwell and Vera platform accelerators over multiple years as demand for AI compute expands in Southeast Asia.
Firmus, an AI infrastructure partner, is working with NVIDIA on a large-scale artificial intelligence data center project in Indonesia that is expected to reach 170,000 GPUs. The initiative is reported to be staged over time, with deployments of NVIDIA’s accelerator and platform offerings planned through 2027 and 2028.
According to the reporting, the facility is intended to use NVIDIA technology built for high-performance AI workloads. The plan includes deploying Vera-Rubin and Grace-Blackwell accelerator platforms, along with the broader “Vera platform” referenced in the announcement. In simple terms, NVIDIA’s accelerator stacks are designed to power model training and high-throughput inference, tasks that require specialized graphics processing units and tightly integrated software.
NVIDIA’s role in such projects typically centers on supplying the GPU-based compute platforms and software ecosystem that system builders can assemble into data centers. For companies like Firmus, the value is in packaging that compute into a location-ready infrastructure that can support customers who need large AI capacity without building everything from scratch.
While the headline GPU count is substantial, the reporting emphasizes a multi-year deployment timeline rather than an immediate “on day one” launch. Spreading installations across 2027 and 2028 can help balance construction schedules, power and cooling readiness, and the availability of the specific NVIDIA hardware and platform configurations being rolled out.
The Vera-Rubin and Grace-Blackwell references point to NVIDIA’s ongoing push to align GPU compute with data center system platforms. In broad terms, Grace-Blackwell refers to a modern platform approach that pairs NVIDIA compute with system-level components optimized for AI datacenter performance, while “Vera” is used here as the umbrella platform framing for the accelerator deployment within the facility.
Indonesia’s push to expand AI-ready infrastructure reflects a wider regional pattern. As organizations in Southeast Asia adopt generative AI and data-intensive machine learning, demand for scalable compute infrastructure has grown, putting pressure on suppliers and integrators to deliver GPU capacity with adequate power, networking, and operational support.
Notably, the reporting available for this update does not spell out financial terms, total project cost, ownership structure, customer contracts, or operating timelines beyond the stated deployment window through 2027 and 2028. It also does not disclose whether the facility capacity is fully committed under specific customer agreements or whether it will be sold or allocated over time.
What to watch next is whether Firmus and NVIDIA provide additional details on the facility’s location, build-out phases, and the specific configuration of the installed accelerators and supporting software. Any future disclosures about customer commitments, contracted GPU utilization, or expansion plans would help clarify how quickly the 170,000-GPU capacity is expected to come online and how it will be monetized.
Why It Matters
- A 170,000-GPU facility indicates rising demand for AI compute infrastructure in Indonesia and the broader region.
- Multi-year deployment suggests system-build considerations such as construction phasing and power and cooling readiness are central to scaling GPU capacity.
- The project highlights NVIDIA’s continued presence in large data center builds that rely on integrated accelerator platforms for AI training and inference.
- Additional disclosures on customer commitments and financial structure would determine how quickly the capacity translates into recurring revenue.
Sources
Key Facts
- Firmus and NVIDIA are reported to be planning a 170,000-GPU AI data center facility in Indonesia.
- The project is scheduled to deploy NVIDIA-related offerings over multiple years, with deployments planned through 2027 and 2028.
- The reporting references deployment of Vera-Rubin and Grace-Blackwell accelerator platforms.
- The plan also references “Vera platform” accelerators as part of the facility’s roll-out.
- The GPU count and deployment window are the primary specifics provided in the available report.
Technology Related
AMD says Instinct AI systems are now operating in Saudi Arabia, highlighting a potential ramp tied to additional data-center power
A recent market report frames AMD’s Instinct deployments in Saudi Arabia as a move from plan to production, and points to how incremental data-center capacity, measured in megawatts, could influence investor expectations.
Salesforce says AI-driven revenue momentum is building as Agentforce adoption spreads
In a recent market update circulated by Yahoo Finance, Salesforce management pointed to expanding use of its AI offerings, including agentic workflows and consumption-style pricing, as the company positions its next growth phase.
Salesforce backs HiBob to bolster workforce AI, and adds a new AgentExchange email tool
Salesforce said it is supporting HR-analytics and talent-workforce platform HiBob as part of efforts to connect enterprise data with “powered AI.” The company also announced an AgentExchange email tool aimed at expanding what business agents can do inside everyday workflows.
EverPass Media expands NFL distribution via multi-year Netflix deal for 2026 slate
EverPass Media says it has added Netflix’s five NFL games for the 2026 season to its NFL distribution offering, including the first-ever Thanksgiving Eve game, plus “NFL Honors.”
Broadcom leans harder into VMware AI with a push aimed at enterprise rivals
Broadcom’s VMware AI push is tied to the latest VCF 9.1 release, as the company’s messaging positions it against Nutanix and Microsoft in hybrid cloud and enterprise AI rollouts.
Yahoo Finance points to “buy zones” for Microsoft, Palantir, Shopify and ServiceNow
A market-readout from Yahoo Finance flagged several software and AI-linked names, including Palantir (PLTR), as trading in or near so-called buy zones. The note is framed as technical or timing-oriented, with limited company-specific detail.
Oracle Shares Fall as Investors Focus on Cash Flow Gap and Rising Borrowing Costs
A reported $23.7 billion cash shortfall over Oracle’s last fiscal year and $43 billion in borrowing are drawing attention to the company’s interest-rate exposure, a factor that can quickly change sentiment when Treasury yields are elevated.
Adobe’s next report faces a split view: Citi still expects a beat, but flags lingering risks
After Adobe lowered its annual revenue outlook, one analyst said the company can still deliver a beat-and-raise in fiscal third-quarter results, even as concerns remain.
Palantir’s commercial growth may overtake government revenue sooner than expected, according to a new market model
A widely watched growth-math forecast argues Palantir’s commercial revenue could surpass its government revenue before 2027, driven by a widening gap in the companies’ growth rates.
Netflix shares face another round of debate after new market commentary, but company keeps details scarce
A recent Yahoo Finance-linked article argues Netflix is not finished telling its story, urging investors to stay cautious until more clarity emerges.