Buy AWS Account

Overview: what a Buy AWS Account listing actually covers

This page lists twenty separate AWS account configurations, spanning entry-level
evaluation tiers through to large parallel compute allocations. They are grouped into
five practical families — trial, raw compute measured in vCPUs, prepaid credit
balances, Amazon Bedrock enabled configurations described by requests per minute, and
configurations that additionally reference Claude model support. Prices run from
$20 for the Free Trial Account to $14,500 for the
100K Credits Account.

The point of the range is that AWS workloads are not one shape. A developer
validating an architecture for a week has almost nothing in common with a team running
sustained inference against a large model. One needs a small, cheap environment to
prove something works. The other needs headroom measured in hundreds of vCPUs and a
request rate that will not throttle under concurrent load. Choosing well means matching
the configuration to the shape of the work rather than buying the largest number
available.

Everything on this page is drawn from the configuration list itself. Where a detail
is not stated in that list — specific regions, particular service quotas, exact
model availability — this page says so plainly rather than guessing. If a
specification matters to your decision, confirm it before ordering via the
contact page.

Key features of the listed AWS account configurations

The points below summarise what distinguishes these options for anyone about to buy AWS account configurations for the first time.

• Twenty distinct tiers so capacity can be matched to workload rather than rounded up.
• Two pricing models — capacity-denominated configurations (vCPU and RPM) and credit-denominated balances.
• Amazon Bedrock enabled options at four separate request-rate and vCPU combinations.
• A Kiro enabled option for agentic development workflows.
• Claude-model configurations that name 4.6, 4.7 and 4.8 support explicitly.
• A $20 entry point so evaluation does not require committing to a large tier.

Available configurations and pricing

The table below is the complete list. Read it as the menu you choose from when you buy AWS account configurations on this page — each row is priced and specified independently, so nothing is bundled or assumed.

The full list, exactly as priced:

Configuration	Family	Typical fit	Price
Free Trial Account	Entry	Evaluation and first-time setup	$20
32vCPUs Account	Compute	Light build and test workloads	$35
128vCPUs Account	Compute	Mid-sized parallel workloads	$100
256vCPUs Account	Compute	Larger batch and training jobs	$150
512vCPUs Account	Compute	High-throughput parallel compute	$250
1K Credits Account	Credits	Short projects billed against credit	$200
5K Credits Account	Credits	Sustained multi-month projects	$950
10K Credits Account	Credits	Team-scale consumption	$1,800
25K Credits Account	Credits	Department or agency scale	$5,500
100K Credits Account	Credits	Long-horizon programme budgets	$14,500
Bedrock Enabled 10 RPM 5vCPUs	Bedrock	Low-rate model experimentation	$70
Bedrock Enabled 50 RPM 5vCPUs	Bedrock	Prototype apps with modest traffic	$120
Bedrock Enabled 10K RPM 32vCPUs	Bedrock	Higher request-rate testing	$150
Bedrock Enabled 10K RPM 256vCPUs	Bedrock	Production-rate inference workloads	$550
Kiro Enabled Account	Kiro	Agentic development workflows	$150
256vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Bedrock + Claude	Claude-model application work	$3,500
384vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Bedrock + Claude	Scaled Claude-model workloads	$4,000
512vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Bedrock + Claude	Concurrent multi-user inference	$5,500
1280vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Bedrock + Claude	Large parallel inference estates	$8,000
1920vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Bedrock + Claude	Maximum listed capacity tier	$9,500

Understanding vCPU, RPM and credit tiers

vCPU-denominated configurations

A vCPU is a virtual CPU — the unit AWS uses to describe how much processing
capacity an instance provides. Broadly, a vCPU corresponds to a thread on the
underlying physical processor. The practical consequence is that vCPU count governs how
much work can proceed in parallel. Doubling vCPUs does not make a single-threaded
task twice as fast; it lets you run roughly twice as many things at once.

That distinction matters when choosing between the 32, 128, 256 and 512 vCPU
configurations. If your workload is a batch that splits cleanly into independent
chunks — data preparation, embedding generation, parallel evaluation runs —
higher vCPU counts translate almost directly into shorter wall-clock time. If it is a
long sequential process, a larger tier mostly buys you the ability to run more of them
side by side.

Configuration	Typical fit	Price
32vCPUs Account	Light build and test workloads	$35
128vCPUs Account	Mid-sized parallel workloads	$100
256vCPUs Account	Larger batch and training jobs	$150
512vCPUs Account	High-throughput parallel compute	$250

AWS documents the vCPU model and how it maps to instance families in the
Amazon EC2 documentation,
which is the authoritative reference for how compute capacity behaves.

RPM-denominated configurations

RPM — requests per minute — describes throughput rather than raw compute.
It is the ceiling on how many API calls a workload can make in a minute before it is
throttled. For model inference this is frequently the binding constraint, because a
single application server can generate far more requests per minute than a low rate
limit permits.

The listed Bedrock configurations span 10 RPM, 50 RPM and 10K RPM. The gap between
those numbers is not incremental, it is categorical. A 10 RPM configuration supports
one request roughly every six seconds — workable for manual experimentation,
unworkable for anything user-facing. A 50 RPM configuration supports a prototype with
light traffic. The 10K RPM configurations are the ones intended for sustained
concurrent use.

Credit-denominated configurations

Credit accounts are denominated in spend rather than capacity. Instead of fixing how
much compute is available at once, they fix how much total consumption is funded. The
1K, 5K, 10K, 25K and 100K tiers differ only in balance.

Configuration	Typical fit	Price
1K Credits Account	Short projects billed against credit	$200
5K Credits Account	Sustained multi-month projects	$950
10K Credits Account	Team-scale consumption	$1,800
25K Credits Account	Department or agency scale	$5,500
100K Credits Account	Long-horizon programme budgets	$14,500

Credits suit workloads with uneven demand. If usage arrives in bursts — a
training run this month, near-idle the next — a credit balance absorbs that
unevenness better than a fixed capacity tier. Conversely, if load is steady and
predictable, a capacity-denominated configuration gives more certainty about what will
be available at peak.

Bedrock enabled configurations

Amazon Bedrock is AWS’s managed service for accessing foundation models through a
single API, without provisioning the underlying inference infrastructure. AWS documents
the service, its supported model providers and its API surface in the
Amazon Bedrock documentation.

Configuration	Typical fit	Price
Bedrock Enabled 10 RPM 5vCPUs	Low-rate model experimentation	$70
Bedrock Enabled 50 RPM 5vCPUs	Prototype apps with modest traffic	$120
Bedrock Enabled 10K RPM 32vCPUs	Higher request-rate testing	$150
Bedrock Enabled 10K RPM 256vCPUs	Production-rate inference workloads	$550

Four configurations are listed as Bedrock enabled. They vary along two axes at once
— request rate and vCPU count — and the two are not independent in practice.
A high request rate with minimal vCPUs will bottleneck on the application side; ample
vCPUs behind a 10 RPM ceiling will sit idle waiting for rate-limit headroom. The
configurations pair the two deliberately: 5 vCPUs at the low rate tiers, 32 vCPUs at
10K RPM, and 256 vCPUs at the highest listed Bedrock tier.

Deeper background on Bedrock, including how the service fits into a wider AWS
architecture, is covered on the Bedrock page.

Claude-model configurations

Five configurations explicitly reference 4.6, 4.7 and 4.8 support, at 256, 384, 512,
1280 and 1920 vCPUs, all at 10K RPM. These are the highest-capacity entries in the list
and are priced accordingly, from $3,500 to $9,500.

Configuration	Typical fit	Price
256vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Claude-model application work	$3,500
384vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Scaled Claude-model workloads	$4,000
512vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Concurrent multi-user inference	$5,500
1280vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Large parallel inference estates	$8,000
1920vCPUs 10K RPM — 4.6, 4.7, 4.8 Supported	Maximum listed capacity tier	$9,500

The relevant question when choosing among them is concurrency, not peak speed. All
five share the same 10K RPM ceiling; what separates them is how much simultaneous work
the account can sustain behind that ceiling. A single-tenant internal tool rarely needs
1920 vCPUs. A platform serving many concurrent users plausibly does.

Kiro enabled configuration

One configuration is listed as Kiro Enabled Account at $150. Kiro is AWS’s
agentic development environment — tooling oriented around an AI agent that can
work through development tasks over extended runs rather than responding to isolated
prompts. Its official documentation is at
kiro.dev, and this
site covers it in more detail on the Kiro page.

This configuration is the right starting point if the intended use is
development-workflow automation specifically, rather than general inference capacity.
It is priced at the same level as the 256vCPUs Account, which makes the choice between
them a question of purpose rather than budget.

Supported AI and cloud workloads

Model inference and application serving

The Bedrock and Claude-model configurations are oriented toward serving inference
requests. The determining factors are the RPM ceiling and enough vCPU capacity to keep
requests flowing without queueing at the application tier.

Batch and parallel processing

Data preparation, embedding generation, evaluation sweeps and other
embarrassingly-parallel work map directly onto the vCPU tiers. These jobs scale nearly
linearly with available parallelism.

Development and integration testing

The Free Trial and 32vCPUs configurations exist for work where the goal is
correctness rather than throughput — wiring up SDKs, validating IAM policies,
confirming an architecture behaves as expected before committing to a larger tier.

Agent-driven development

The Kiro configuration targets workflows where an agent executes multi-step
development tasks, which have a different resource profile from request-response
inference.

Who should buy AWS account configurations

• Individual developers evaluating AWS AI services — Free Trial Account ($20) or 32vCPUs Account ($35).
• Small teams building a prototype — Bedrock Enabled 50 RPM 5vCPUs ($120) or Kiro Enabled Account ($150).
• Teams moving a prototype toward production traffic — Bedrock Enabled 10K RPM 32vCPUs ($150) or 256vCPUs ($550).
• Applications serving concurrent users against Claude models — the 256 to 512 vCPU Claude configurations ($3,500–$5,500).
• Large parallel inference estates — 1280 or 1920 vCPU configurations ($8,000–$9,500).
• Budget-driven rather than capacity-driven programmes — the credit tiers, from 1K ($200) to 100K ($14,500).

Use cases

Retrieval-augmented generation

RAG systems have two distinct resource profiles: an indexing phase that is heavily
parallel, and a serving phase that is rate-limited. Teams often find the indexing phase
suits a high-vCPU configuration while steady-state serving is governed by the RPM
ceiling. A configuration such as Bedrock Enabled 10K RPM 256vCPUs covers both without
switching environments.

Document and content processing pipelines

Large-volume extraction, classification and summarisation work splits cleanly across
workers. Wall-clock time falls roughly in proportion to vCPU count, which makes the
128, 256 and 512 vCPU tiers straightforward to reason about.

Internal developer tooling

Assistants and code-review tooling used inside an organisation typically generate
modest sustained request rates with sharp peaks at working-hours boundaries. Headroom
matters more than average capacity here, which is an argument for a 10K RPM
configuration even when average load looks small.

Evaluation and benchmarking

Comparing model behaviour across prompts, versions or parameters is bounded by how
many evaluations can run concurrently. This is the clearest case for the high-vCPU
Claude configurations, where the work is entirely parallel.

How to size a configuration before you buy

Most sizing mistakes come from estimating capacity in the abstract instead of
measuring the workload. Three measurements settle the question in most cases.

Measure your actual request rate, not your user count

Request rate and user count are only loosely related. A hundred users clicking
occasionally may generate fewer requests per minute than ten users running an automated
workflow. What matters is calls per minute at peak, including retries and any background
jobs. If an existing system already runs, instrument it for a week and take the busiest
minute rather than the average — rate limits bind at peaks, not at means.

Once you have that number, add headroom. A configuration sized exactly to observed
peak will throttle the first time traffic exceeds its previous maximum. The step from
50 RPM to 10K RPM in the listed configurations is large enough that most workloads with
any real concurrency belong above the 50 RPM tier.

Determine whether the work is parallel

vCPU count only helps work that can be split. A useful test: if you doubled the
number of machines, would the job finish in roughly half the time? For batch
processing, embedding generation and evaluation sweeps the answer is usually yes, and
higher vCPU tiers pay off directly. For a sequential pipeline where each step depends on
the previous one, extra vCPUs let you run more pipelines side by side but will not speed
up any single run.

Decide whether demand is steady or bursty

Steady demand favours capacity-denominated configurations, because you can size to a
known peak and keep it. Bursty demand — a heavy month followed by a quiet one
— favours credit tiers, because unused capacity in a fixed tier is simply wasted
whereas an unspent credit balance is not. Teams whose usage is genuinely unpredictable
often start on a credit tier precisely to discover their real consumption pattern before
committing to a capacity tier.

How AWS accounts are structured

Understanding a few structural facts about AWS makes the configuration list easier to
interpret, and they apply to any AWS account regardless of how it was obtained.

Regions and availability

AWS operates in discrete geographic regions, and service availability differs between
them — not every service or model is offered everywhere, and quotas are tracked
per region. AWS maintains a current matrix in its
regional services list.
The configurations on this page are described by capacity and rate; they do not state
regions, so confirm regional requirements separately if your project has them.

Service quotas

AWS enforces per-account, per-region limits called service quotas, documented in the
Service Quotas documentation.
Some are adjustable on request, others are hard ceilings. This is worth knowing because
quotas are a distinct constraint from the capacity described in a configuration name: an
account can have substantial vCPU capacity and still be limited by a quota on a
particular service.

Identity and access management

Access is controlled through IAM — users, roles and policies that determine what
any given credential can do. AWS’s own guidance is to avoid routine use of root
credentials and to operate through scoped IAM identities instead, with multi-factor
authentication enabled. The
IAM security best practices
are the reference here, and following them is the single highest-value thing to do on
any new account before putting real work on it.

Billing and cost visibility

Consumption-based billing means costs track usage rather than a flat subscription.
For credit-denominated configurations the practical implication is that a runaway job
consumes balance quickly. Setting up budget alerts early, using AWS Budgets, is
inexpensive insurance against discovering a problem only when the balance is gone.

Configuration comparison

The practical decision usually reduces to three questions.

Is the constraint throughput or total spend? If you need a guaranteed
request rate at peak, choose a capacity configuration. If consumption is uneven and the
budget is the real limit, choose a credit tier.

Does the workload need Bedrock specifically? The plain vCPU tiers and
the Bedrock enabled tiers are priced very differently at similar vCPU counts —
compare the 256vCPUs Account at $150 against Bedrock Enabled 10K RPM 256vCPUs at $550.
The difference reflects the Bedrock enablement and request-rate ceiling, not the raw
compute.

Is Claude model support a requirement? Five configurations name 4.6,
4.7 and 4.8 explicitly. If those models are required, the choice is among those five and
the question becomes how much concurrency to buy.

How the ordering process works

Select the configuration you need from the selector at the top of this page, add it
to the cart, and complete checkout. Payment is handled through the configured
cryptocurrency gateway at checkout. The checkout collects only a name and an email
address — no billing address is requested, because these are non-physical
products.

If you are unsure which tier fits, ask before ordering rather than after. Describe
the workload — expected concurrency, whether Bedrock is required, whether specific
model support matters — and it is usually possible to narrow the list to two or
three candidates quickly.

Important considerations before ordering

Confirm the specification you depend on. The configuration names on this
page describe vCPU counts, request rates, credit balances and stated model support. They
do not describe regions, individual service quotas, or availability of any particular
model beyond what is named. If your project depends on a detail not listed, confirm it
in advance.

Cloud provider terms apply to account usage. AWS publishes its terms in the
AWS Customer Agreement
and its acceptable-use rules in the
AWS Acceptable Use Policy.
Reviewing both before purchase is worthwhile, since they govern what any AWS account may
be used for and under what conditions access may be affected.

Capacity is not the same as performance. A large vCPU allocation does not
improve a workload that cannot use parallelism, and a high RPM ceiling does not help an
application that cannot generate that many requests. Match the tier to the actual
bottleneck.

Prices are per configuration. Each listed price corresponds to the named
configuration exactly as shown. Nothing on this page implies a discount, bundle or
recurring arrangement that is not stated.

Frequently asked questions

Which configuration should I choose?

Match the configuration to the workload rather than to the price. Most people who buy AWS account configurations start from the smallest option that covers their current need and move up once real usage is measured, which avoids paying for headroom that never gets used. The comparison sections above set out the trade-offs; the FAQ page covers ordering questions.

Is it safe to buy AWS account configurations listed here?

Every configuration on this page states its specification and price exactly as supplied, and nothing is implied beyond what is written. Anything your project depends on that is not named should be confirmed on the contact page before you buy AWS account configurations.

What is the difference between a vCPU tier and a credit tier?

A vCPU tier fixes how much compute capacity is available at any moment. A credit tier
fixes how much total consumption is funded, leaving the capacity mix up to you. Choose
capacity when peak throughput matters; choose credits when the budget is the constraint
and demand is uneven.

Which configurations include Amazon Bedrock?

Four are named Bedrock Enabled: 10 RPM 5vCPUs ($70), 50 RPM 5vCPUs ($120), 10K RPM
32vCPUs ($150) and 10K RPM 256vCPUs ($550). The five Claude-model configurations at
$3,500 to $9,500 also reference Bedrock-served models.

What does “4.6, 4.7, 4.8 Supported” mean?

It refers to the Claude model versions named in those configuration titles. Five
configurations carry that designation. No claim is made here about models beyond the
ones named.

Is 10K RPM always better than 50 RPM?

Only if your application can generate that volume. A prototype making a few requests
a minute gains nothing from a higher ceiling. Pay for rate headroom when concurrency is
real, not anticipated.

How do I choose between the 1280 and 1920 vCPU configurations?

Both run at 10K RPM, so the difference is concurrency, not request rate. If your
current workload saturates 1280 vCPUs, the larger tier gives room to grow; if it does
not, the extra capacity will sit idle.

Can I move to a different configuration later?

Each configuration is sold as its own item at its own price. If your requirements
change, discuss it via the contact page before ordering
something new.

What payment methods are accepted?

Checkout uses the cryptocurrency payment gateway configured on this site. It appears
automatically at the payment step; no separate arrangement is needed.

Where can I read more about the underlying services?

This site covers the services in more depth on the
AWS AI page, the Bedrock page and the
Kiro page. Common questions are collected on the
FAQ page. Authoritative technical detail is in the
official AWS documentation.

What you get when you buy AWS account

Every listing on this page is a specific, named configuration rather than an open-ended arrangement. When you buy AWS account here you are choosing one of the 20 configurations described above, at the stated price $20.00 – $14,500.00, with the specification exactly as written.

• The exact configuration named in the selector, with no substitutions.
• The price shown, with no added fees at checkout.
• A checkout that asks only for a name and an email address.
• Direct contact before ordering if you are unsure which option fits.

If you are comparing providers rather than configurations, the Cloud Accounts and Cloud Infrastructure categories list the alternatives side by side. If you already know what you need, you can buy AWS account configurations directly from the selector at the top of this page.