Buy Now, Buy Later, or Something Else?

If you’ve priced a new server, laptop, or even a memory upgrade recently, you’ve probably had the same reaction many IT leaders are having right now: “How did this get so expensive so fast?”

The short answer is AI.

The longer answer is that the explosive demand for AI infrastructure is fundamentally reshaping the hardware market. Massive datacenters filled with GPUs, high‑density servers, and unprecedented amounts of memory are being built at a scale we’ve never seen before. And that demand is colliding with a supply chain that simply cannot expand overnight.

So the big question organizations are asking today is simple—but the answer is not:

Do we buy now, buy later, or do something else entirely?

The Case for Buying Now

The “buy now” argument is built on a simple reality: prices are not expected to come down anytime soon.

Industry forecasts are pointing to continued price increases for at least the next three years, with some components—especially memory—seeing steady month‑over‑month growth. A commonly cited expectation across the industry is roughly 1% price increases per month, compounding quickly over time.

As one Dell executive recently put it:

“The cheapest day to buy hardware is today.”

That statement may sound dramatic, but it reflects what we’re seeing in real quotes every day. Waiting even a few weeks can mean paying meaningfully more for the same configuration. For organizations with known refresh cycles, expiring warranties, or growth plans that can’t be delayed, buying now can be the least risky option—even if prices feel uncomfortable.

The Case for Buying Later

On the other side of the debate is the hope that supply and demand will eventually rebalance.

There is no question that more manufacturing capacity is coming online. Memory producers, chip manufacturers, and hardware vendors are all investing heavily to increase output. In theory, more supply should help stabilize pricing.

However, the challenge is timing.

The largest technology companies in the world—Amazon, Microsoft, Google, Apple, and Meta—are investing tens of billions of dollars each into AI datacenters. These companies are not just customers; they are effectively absorbing massive portions of global hardware production.

Most analysts expect supply to get tighter, not better, over the next 24–36 months before meaningful relief appears. That means “buy later” may actually mean buying several years from now, and accepting higher prices in the meantime.

For organizations that can safely extend existing hardware or delay non‑critical projects, waiting can make sense—but it’s important to do so with eyes wide open.

Something Else: Rethinking Hardware Altogether

For many organizations, the real answer isn’t buying now or later—it’s buying differently.

One increasingly common alternative is moving server infrastructure to the cloud, particularly platforms like Microsoft Azure. By doing so, organizations can avoid large capital hardware purchases altogether. There are no servers to buy, refresh, or maintain—and in theory, no hardware purchases ever again.

That said, the cloud is not immune to these same market pressures. As underlying hardware costs rise, cloud pricing will rise as well. The difference is that those costs show up as predictable operational expenses rather than large, upfront capital investments.

On the end‑user side, there are fewer immediate alternatives to physical desktops and laptops—unless those, too, move to the cloud. More organizations are exploring fully virtual desktop infrastructures, where users access secure desktops hosted in the cloud. While this approach isn’t cheap, it replaces irregular, large capital expenditures with a more predictable monthly operating cost.