In the shadow of a global memory shortage with no near-term resolution, Nvidia has quietly redefined what it costs to participate in the AI hardware frontier. The DGX Spark's 128GB model, once priced at $3,999, now commands $6,950—a 74 percent climb that mirrors a deeper truth about the infrastructure age: the materials that make intelligence possible are finite, and their scarcity shapes who can afford to think at scale. With Micron projecting tight DRAM supply through 2028, this moment is less a pricing decision than a structural reckoning with the physical limits of the AI boom.
Nvidia Raises DGX Spark Prices 74%, Launches 64GB Model at $4,999
Memory shortage expected through 2028 gives Nvidia little reason to cut prices.
Mark
So Nvidia just doubled the price of the DGX Spark in under a year. Is this a supply problem or a demand problem?
Mimi
It's supply. Micron, which makes the memory inside these things, says DRAM will be tight through 2028. That's not a short-term blip—that's structural.
Luke
But we should be careful here. Nvidia cites memory supply, but we don't have Nvidia's internal cost data. We know Micron warned about supply, but we don't know exactly how much of that 74 percent increase is actual cost versus margin expansion.
Mimi
Fair point. But the February bump to $4,699 happened before this one, so there's a pattern of escalation tied to the same supply story.
Mark
Why introduce the 64GB model now instead of just raising prices on the 128GB?
Mimi
Market segmentation. Some customers don't need 128GB. A smaller model at $4,999 captures price-sensitive buyers who would have walked away at $6,950.
Luke
That's smart business, but it also means Nvidia is confident enough in demand that they can afford to segment upward. If demand were soft, they'd be discounting, not introducing premium tiers.
Mark
What happens if Micron's timeline slips and supply stays tight past 2028?
Mimi
Then these prices probably stay elevated. Memory fabrication takes years to build out capacity.
Luke
And if supply loosens faster than expected, Nvidia could face pressure to cut prices. But that's speculative. Right now we only know what Micron said and what Nvidia is charging.
O Pulso
- Nvidia's flagship personal AI accelerator has nearly doubled in price within a year, catching existing and prospective customers in a steep and accelerating cost curve.
- A global DRAM shortage—structural in nature and years in the making—is the engine behind the pressure, with Micron warning supply will remain constrained through 2028.
- To soften the blow and segment the market, Nvidia introduced a 64GB variant at $4,999, available October 23 through major OEMs including Dell, HP, ASUS, and MSI.
- Two 64GB units can be clustered via ConnectX-7 to reach 128GB capacity with a claimed 1.7x performance gain, offering a modular path for buyers unwilling to absorb the full premium.
- Nvidia's confidence in pushing prices higher signals that demand among AI infrastructure builders remains resilient—customers are paying what the market demands rather than waiting for relief.
In the shadow of a global memory shortage with no near-term resolution, Nvidia has quietly redefined what it costs to participate in the AI hardware frontier. The DGX Spark's 128GB model, once priced at $3,999, now commands $6,950—a 74 percent climb that mirrors a deeper truth about the infrastructure age: the materials that make intelligence possible are finite, and their scarcity shapes who can afford to think at scale. With Micron projecting tight DRAM supply through 2028, this moment is less a pricing decision than a structural reckoning with the physical limits of the AI boom.
Nvidia's DGX Spark has grown dramatically more expensive in under a year. The 128GB model launched at $3,999, climbed to $4,699 in February, and now sits at $6,950—a 74 percent total increase driven by a global memory shortage that shows no sign of easing. Micron, one of the world's largest DRAM manufacturers, has warned that supply will remain tight through 2028, a timeline that gives hardware makers little reason to expect relief. Memory fabrication plants take years and billions of dollars to build; the shortage is structural, not cyclical.
Rather than simply raising prices, Nvidia used the moment to reshape its product lineup. A new 64GB variant, priced at $4,999 and available October 23 through Acer, ASUS, Dell, Gigabyte, HP, and MSI, creates a lower entry point for customers with more modest computational needs. Both models share the same GB10 Grace Blackwell Superchip—a 20-core Arm CPU paired with 273 GB/s of unified memory bandwidth—differing only in capacity. The 64GB unit supports open-source AI models up to 100 billion parameters, and two units can be clustered via ConnectX-7 ports to pool memory and achieve what Nvidia claims is 1.7 times the performance of a single system.
For those already invested in the DGX Spark ecosystem, the escalation is stark. A customer who bought at launch now faces a 74 percent premium for equivalent hardware. That Nvidia is willing to push pricing this aggressively reflects a broader confidence: that the builders of AI infrastructure will absorb the cost rather than wait for cheaper alternatives. In an era defined by the race to deploy AI at scale, the price of memory has become the price of admission.
Nvidia's DGX Spark has become significantly more expensive in less than a year. The 128GB model, which debuted at $3,999, now carries a price tag of $6,950—a 74 percent increase that reflects the company's response to a global memory shortage that shows no signs of easing soon. An intermediate price hike in February had already lifted the Founders Edition to $4,699, a jump of 18 percent, but the latest move represents a far steeper climb.
The company is not simply raising prices on existing products. Alongside the price increase, Nvidia introduced a new 64GB variant priced at $4,999, available starting October 23 through major manufacturers including Acer, ASUS, Dell, Gigabyte, HP, and MSI. The move appears designed to segment the market—offering a lower-cost entry point for customers with smaller computational needs while maintaining the premium pricing on the higher-capacity model.
Both versions share the same core processor: the GB10 Grace Blackwell Superchip, which pairs a 20-core Arm CPU with 273 gigabytes per second of unified memory bandwidth. The distinction lies purely in capacity. The 64GB unit can handle open-source AI models up to 100 billion parameters, a meaningful but not massive scale. For larger workloads, Nvidia's architecture allows two 64GB systems to cluster together via ConnectX-7 ports, pooling their memory to reach 128GB and delivering what the company claims is 1.7 times the performance of a single unit.
The pricing pressure stems from a genuine supply constraint. Micron, one of the world's largest memory manufacturers, has publicly warned that DRAM supply will remain tight through 2028. That timeline gives Nvidia and other hardware makers little reason to expect relief in the near term. Memory is not a commodity that can be quickly ramped up; fabrication plants take years to build and billions of dollars to construct. The shortage is structural, not temporary, and Nvidia's pricing reflects that reality.
For customers already committed to the DGX Spark ecosystem, the price increases represent a significant cost escalation. A customer who purchased a 128GB unit at launch would now pay 74 percent more for the same hardware. Even accounting for the new 64GB option, the entry price into Nvidia's latest AI accelerator platform has risen substantially. The company's willingness to push prices higher suggests confidence in demand—that customers building AI infrastructure will pay what the market will bear rather than wait for cheaper alternatives or competing products.
Citações Notáveis
Two 64GB systems clustered together deliver roughly 1.7 times the performance of a single Spark— Nvidia's performance benchmark
Quer a matéria completa?
Leia o original em AI Weekly ↗