Written 11 October 2026.
Every few months a wave of posts tells you to re-audit your hosting bill. The reason behind this quarter's wave sits upstream of any hosting company, in the price of the memory chips that servers are built from. The primary document is a 30 September 2026 forecast from TrendForce, the research firm that tracks memory contract pricing. It is short, it is specific, and it is worth reading before you read anyone's commentary on it, including this.
What the forecast actually says
Two sentences carry the numbers. On DRAM: "Conventional DRAM contract prices are projected to grow 10–15% QoQ in 4Q26." On flash storage: "Overall NAND Flash contract prices are projected to increase 15–20% QoQ." The cause TrendForce gives is demand from AI servers, with suppliers directing capacity there and leaving less for everything else.
Three things to keep straight, because they get blurred within a day of any forecast like this:
- It is a forecast, not a price list. "Projected" is the operative word. These are expected movements in contract prices between chip makers and their large customers for the fourth quarter of 2026.
- It is about memory chips, not hosting. Nothing in the document is a hosting price. A 10–15% move in conventional DRAM is not a 10–15% move in anyone's monthly bill.
- The server figure is not there. On server DRAM specifically, the document says the market is "expected to remain undersupplied in 4Q26" and that "increases for some suppliers are expected to lag the market average" because of long-term agreements. It gives no percentage for server DRAM in this quarter. If you see one attached to this forecast, it did not come from this forecast.
The document also does not say when the trend ends. Treat any date you read for that as someone's guess.
Why a chip forecast reaches your hosting bill at all
A hosting provider is a buyer of servers. When a provider adds capacity, it buys machines whose cost is partly memory and storage, and if those parts cost more, the machines cost more. Over time that pressure has to go somewhere: into the provider's margin, or into prices. That is a structural statement about the industry, not a prediction about any particular company. As of the day this was written, no hosting provider we track has announced a price change tied to this forecast, and this article will not guess at one.
So the useful question is not "will my host raise prices", which nobody outside that host can answer. It is: if a provider's costs rise, which line on my bill is the one that moves? The answer depends on how the provider charges, and there are two common shapes.
Shape one: the meter
On a usage-priced platform you pay for what you consume: requests, compute-seconds, gigabytes transferred, invocations, seats. Hardware cost reaches you through the meter in two ways. The first is volume: a busier month is a bigger bill whether or not anything changed on the provider's side. The second is the rate under the meter, which the provider sets and can revise. Both can move at once, and the second often moves quietly, as a line in a pricing-page update.
The strength of this shape is that a small app pays a small bill. The weakness is that the bill is a function of two variables you do not control: your users' behaviour and your provider's rate card. You find out what a good month costs after it happens.
Shape two: the box
On a plan priced by server size you choose a fixed amount of RAM, CPU and storage, with a bandwidth allowance, and pay a fixed monthly price for it. The thing you bought is the box. Traffic inside the allowance does not change the bill, and if a provider's hardware costs rise, the only place that can reach you is the plan price itself, which is a single published number you can see change. Going past the allowance is handled as a limit on the plan, not as a surcharge that appears on the invoice.
The strength of this shape is predictability: the same bill every month, and one number to watch. The weakness is the inverse of the meter's: a quiet app pays for capacity it is not using, and a sudden spike runs into the plan's ceiling rather than being metered for a fee.
DeployBase prices by server size. Plans differ by RAM, vCPU, storage and bandwidth allowance, and there are no per-request or bandwidth charges. That describes the shape of the bill, not where the number on it goes; pricing announcements are made on the pricing page, not in blog posts.
How to read your own bill against this
- Find the meter. Open your last three invoices and mark every line whose amount varied month to month. Those are the lines a cost increase can reach without anyone telling you.
- Find the rate. For each metered line, note the unit price. Save the pricing page. If the rate changes, you want to be able to prove when.
- Price your worst month, not your average. Take your busiest month's usage and multiply by today's rates. That is the number a fixed plan should be compared against, because that is the month the meter hurts.
- Know the ceiling on a fixed plan. If you are on a box, know its bandwidth allowance and what happens at the edge of it, so a good month is a known quantity rather than a surprise either way.
- Re-check quarterly. Memory contract prices move quarter by quarter, and so do pricing pages. A fifteen-minute look at your bill every quarter is cheaper than discovering a rate change six months after it happened.
None of this tells you which shape is right. A metered platform that charges you two dollars a month for a side project is the right answer for that project. The point is only that when input costs rise, as this forecast says they will for memory in the fourth quarter of 2026, the two shapes pass that cost to you through different doors, and you should know which door your bill has.
Source, read 11 October 2026: TrendForce, "AI Server Demand Sustains Memory Contract Price Increases in 4Q26, While Consumer-Side Pressure Persists", 30 September 2026 (trendforce.com). DeployBase plan facts are from the live pricing page on the same date.