首屏：10 张行业轮播 + 打字机特效 + 自定义光标 → 单一定位句 + B300 机房示意 + 两个按钮。现网：Building Global AI Inference Service Infrastructure / XMAX GROUP SYSTEM
NewDecember 2026: three platform releases
From compute
to tokens.
XMax AI runs the full production line behind enterprise AI: GPU capacity, token production, model services and industry applications. Part of XMax Inc. (Nasdaq: XMAX).
Publicly listed parent
Models in production
Export-control review on every deployment
证据条：现网的 30+ regions / 100+ solutions / 99.9%（无依据）已在上一版删除。四条事实全部有来源：上市主体、$120M 协议与 64 张 B300、四层产线、模型服务在产。
- Nasdaq: XMAXParent XMax Inc.
- 64 × B300US$120M, Korea, contracted
- Four layersOne operator
- Models live5.21B tokens / month
- Korea, Australia, USContracted and planned
四层产线：据 2026-11 演示稿，原三层（Compute / Inference / Enterprise AI）改为四层：L1 NeoCloud → L2 Token Factory → L3 MaaS → L4 Application Factory。名称与顺序在首页、大菜单、产品页三处必须一致（PRD 验收标准第 3 条）。
One production line, four layers
From the data hall to the application, four layers on one line: capacity at the bottom, turned into tokens, then APIs, then deliverable applications. Cost, performance and accountability stay with one operator.
Data halls, power and coolingContracted facilities
NeoCloud
Managed GPU capacity with the network, storage and operations around it. Buy it as instances, bare metal, dedicated clusters or reserved capacity.
Compute
Taking reservationsGPU instances and bare metal, dedicated compute, multi-node clusters, Blackwell Ultra and the Rubin roadmap.
Network
With capacityScale-up and scale-out fabrics, GPU interconnection, NVLink, RDMA and high-bandwidth networking.
Storage
With capacityDistributed storage, model weights and data access, high-performance storage orchestration, GPU Direct Storage.
Managed operations
AvailableProvisioning, resource monitoring, scheduling, deployment and fault detection.
Token Factory
The production layer that turns GPU capacity into measurable, verifiable tokens: capacity and models in; serving, scheduling, optimization and metering in the middle; API-ready tokens out.
Delivery models
AvailableServerless endpoints, dedicated capacity, batch processing, accelerated models and customer-owned models.
Availability targets
Per agreementAt least 99.5% for serverless and 99.9% for dedicated capacity.
MaaS
Unified model access, routing, usage metering and enterprise integration. One key, one bill, with quotas and budgets set per team.
Model access
AvailableAccess models through one integration layer instead of a separate contract for each provider.
API and routing
AvailableAuthentication, request routing and gateway policy, with planned expansion on the roadmap.
Metering and billing
AvailableToken-level metering that supports usage-based commercial agreements and billing reconciliation.
Application Factory
Coding agents that turn a requirement into a deployable industry application, with people holding the quality gates. Five stages: requirements, architecture, coding, testing, deployment and operations.
Human oversight
Every projectRequirements approval, code review, industry-expert validation and compliance review.
Reusable assets
In developmentComponent library, industry packages, specification templates, quality framework and delivery SOPs.
GPU 部署管线：新增（PRD 5.1 第 5 块）。严格分开"已签约"与"计划交付"，不能合并成"我们有多少卡"。韩国一行与 $120M 协议是同一条事实。
Where the GPUs are
Contracted capacity and planned deliveries are listed separately. Planned dates are forward-looking and may change.
NeoCloud details| Region | GPU and quantity | Status | Target date |
|---|---|---|---|
| South Korea | 64 × NVIDIA B300US$120M agreement | Contracted | Per agreement |
| Australia | 32 × NVIDIA B300 | Planned delivery | 31 Dec 2026 |
| United States | 4 × NVIDIA GB300 | Planned delivery | 30 Mar 2027 |
Capacity comes online in batches under agreements and delivery plans. Forward-looking; see the notice in the footer.
商业进展：新增（PRD 5.1 第 6 块）。全部数字等 IR 逐条放行，未放行的整块隐藏，不留空（PRD F-12）。
Commercial traction
Recognized revenue, signed agreements and pipeline. Every figure carries its source and cutoff date. Pipeline is opportunity, not recognized revenue.
See investor details$1.10M
AI revenue · H1 2026
Pending IR wording
$4.5M
AI revenue · Q3 2026
Pending IR wording
Up to $25M
AI API agreement value
Framework ceiling, not recognized revenue
Public figures follow IR-approved wording. Figures not yet cleared are hidden rather than shown blank. The prospective pipeline is internal and is not shown here.
发布日历：新增（PRD 5.1 第 7 块 + 5.9）。三个 2026-12 发布是目前最强的"不是 PPT 公司"证据，放首页中段。
What ships next
Three platform releases land together in December 2026, then a platform upgrade each quarter.
Full roadmapDecember 2026
NeoCloud 1.0
The first GA release of GPU capacity, network, storage and managed operations.
PlannedDecember 2026
MaaS 2.0
Expanded model access, routing and enterprise integration.
PlannedDecember 2026
Token Factory 1.0
The production line that turns capacity into metered tokens goes live.
PlannedB300 机房：已确认 $120M 协议与韩国 64 张 B300 是同一件事，合并为一条事实（PRD D-02）。服务日期演示稿未给，官网不给。
The B300 facility
Contracted dedicated capacity: a US$120M agreement, 64 NVIDIA B300 GPUs, in South Korea. Built for large training runs and high-throughput inference.
Full specifications- Agreement value
- US$120MContracted
- Accelerator
- NVIDIA HGX B300Blackwell Ultra, 288 GB HBM3e per GPU
- GPU count
- 64B300 in Korea
- Location
- South KoreaService date per agreement, not public
模型目录：新增。现网公开页面没有任何模型。首发 4 个，版本按平台实际上架核对；价格不上网，统一询价。
Open models, served on our own GPUs
The launch catalog covers the open models enterprises ask for most. Every model is available as a shared API, and as a dedicated endpoint on request.
Browse the catalog合作流程：新增，回应顾问"工作流用可视化/交互方式呈现"。替换现网九个行业页里大段 workflow/review 文字。
From first call to production
Every engagement follows the same five steps. Select a step to see what happens and what you get.
What happens
What you get
解决方案：现网首页 9 张行业轮播 + 9 个独立页，大量否定式免责长句。首页只留 6 个入口，完整 9 个在 Solutions 页。保留哪几个待 D5 拍板。
Where customers put it to work
The same four-layer line serves very different industries. These are the workflows we are asked about most.
All solutions路线图：新增（老板点名 vision / goals / roadmap）。三阶段对外版，时间窗待 IR；内部"12–24 个月万卡级"不上网。
Where we are going
Every company should be able to run frontier-class AI on infrastructure it can trust and afford. Here is how we get there.
Timelines are forward-looking and may change. See the notice at the bottom of this page.
合作伙伴与合作方式：按演示稿第 13 页补三类伙伴与四种合作方式；已具名伙伴沿用 v0.7 的分组与措辞（只写已签内容，不升级关系等级）。
How to work with us
Work with us at the infrastructure, inference, API or application layer, in four ways: GPU rental, reserved capacity, token consumption and application development.
PartnersInfrastructure
- GPU supply
- Data centers
- Connectivity
- Regional deployment
Technology
- Models
- Inference optimization
- AI services
- Integrations
Commercial
- Enterprise distribution
- API resale
- Industry solutions
GPU rentalCapacity by the hour or the term
Reserved capacityNodes and clusters locked on long terms
Token consumptionModel services billed on usage
Application developmentIndustry applications, delivered and run
信任：现网首页没有任何信任元素。沿用现网 Trust 页"四步部署审查"（原意保留，措辞简化），加主体事实表。
Every deployment is documented before it runs
Advanced GPUs come with real obligations. We record who uses the hardware, where and for what, and we check it before anything goes into production.
Visit the Trust Center新闻：现网首页 3 条 2025 年新闻（早于公司成立，须下线），新闻页本身是 "Coming Soon"。改为与新闻稿同步的真实新闻，B300 置顶。
Latest news
Announcements from XMAX AI and XMax Inc., with links to the original press releases.
All news转化区：现网 "Build the Future of Global AI Inference" + 无反馈表单（顾问第 9 条）。改为四条需求入口，直达预选好类型的表单；地址移到页脚。
Plan your capacity with us
Tell us what you want to run. A solutions engineer will reply with options, timing and next steps.
Contact salesPrefer email? Write to info@xmax.com
The B300 facility
A dedicated NVIDIA HGX B300 facility: a US$120M agreement, 64 B300 GPUs, in South Korea, contracted. Specifications and dates follow the agreement and public announcements.
- Accelerator
- NVIDIA HGX B300Blackwell Ultra, 288 GB HBM3e per GPU
- GPU count
- 64NVIDIA B300
- IT power
- [X] MWCritical IT load
- Compute fabric
- [800G InfiniBand]Rail-optimized [TBC]
- Cooling
- [液冷]Direct-to-chip [TBC]
- Location
- South KoreaOperator [如可披露]
- Service date
- Per agreementNot public
- Technology partner
- NVIDIA[计划与等级名称]
Where our capacity comes from
We combine facilities under our own contracts with partner data centers and cloud capacity, then run all of it through one scheduling and billing layer.
Contracted facilities
Dedicated GPU capacity under long-term agreements, starting with the B300 facility.
Partner data centers
GPU capacity hosted with partners such as KT in Korea.
Cloud capacity
Elastic cloud resources for burst demand, platform services and global access.
原 /aws 页：删除 39 / 123 / 750+（AWS 自身数据，非 XMAX）、"joint go-to-market with AWS"、无署名引语块（顾问第 7 条）。只写已签协议范围。
Cloud layer
The XMAX AI platform runs its API front end, control plane and elastic overflow on AWS, under a platform agreement signed in April 2026. Identity, encryption keys and audit logging use the cloud provider's native controls.
Platform software
One software layer turns the hardware into services you can call, with the same controls across every workload.
| Component | What it does |
|---|---|
| Scheduler | Places training jobs and model replicas on the right capacity |
| Model gateway | One API across models, with routing rules, quotas and usage logs |
| Metering and billing | Per-token and per-GPU-hour usage, reconciled to each agreement |
| Controls | Access policies, audit logs and data-retention settings |
Deployment review
Every deployment is documented and approved before it runs, including an export-control review of hardware destinations and end users.
How we review deploymentsReserve B300 capacity
Tell us your GPU count, term and start date.
产品页：原 Compute 页升级为 NeoCloud（L1），按演示稿第 4 页补网络、存储、托管运维三块；GPU 管线表单独成节，严格区分已签约与计划交付。
- Home
- Products
- NeoCloud
NeoCloud
Managed GPU capacity with the network, storage and operations around it. Buy it as instances, bare metal, dedicated clusters or reserved capacity.
Taking reservations
Compute
GPU instances and bare metal, dedicated compute and multi-node clusters, on a hardware roadmap that covers Blackwell Ultra and Rubin.
- Forms
- Instances, bare metal, multi-node clusters
- Hardware
- NVIDIA HGX B300Blackwell Ultra
With capacity
Network
Scale-up and scale-out fabrics, GPU interconnection, NVLink, RDMA and high-bandwidth connectivity.
With capacity
Storage
Distributed storage, model weights and data access, high-performance storage orchestration and GPU Direct Storage.
Available
Managed operations
Provisioning, resource monitoring, scheduling, deployment and fault detection, run by our infrastructure team.
GPU deployment pipeline
Contracted capacity and planned deliveries are listed separately. Planned dates are forward-looking.
| Region | GPU and quantity | Status | Target date |
|---|---|---|---|
| South Korea | 64 × NVIDIA B300US$120M agreement | Contracted | Per agreement |
| Australia | 32 × NVIDIA B300 | Planned delivery | 31 Dec 2026 |
| United States | 4 × NVIDIA GB300 | Planned delivery | 30 Mar 2027 |
Facility specifications are on the B300 facility page.
Reserve B300 capacity
Tell us your GPU count, term and start date.
新页面：Token Factory（L2）在旧版官网不存在，据演示稿第 6、7 页新建。不写 12B tokens/日、85% 利用率、¥2.5 成本目标（PRD 第六节：内部），也不放与 AWS Bedrock 或号池的对比表（第七节）。
- Home
- Products
- Token Factory
Token Factory
The production layer that turns GPU capacity into measurable, verifiable, enterprise-grade tokens. Capacity and models go in; serving, scheduling, optimization and metering happen in the middle; API-ready tokens come out.
Available
Four stages
GPU capacity and models, then inference and scheduling, then optimization and metering, then API-ready token output. Each stage is measured so unit cost keeps falling.
Available
Delivery models
Five forms, chosen by workload: serverless endpoints, dedicated capacity, batch processing, accelerated models and customer-owned models.
Per agreement
Availability targets
At least 99.5% for serverless endpoints and 99.9% for dedicated capacity. Targets and service credits are set in each agreement.
Ongoing
How we measure
We optimize for useful output rather than GPU occupancy: throughput, time to first token, availability and cost per million tokens, measured together.
Start with the API
Start small on the shared API, then move to dedicated capacity as traffic grows.
产品页：原 Inference 页升级为 MaaS（L3）。禁用词：Token Proxy、中转、号池一律不出现；模型接入只写服务能力，不写供给方式。
- Home
- Products
- MaaS
MaaS
Unified model access, routing, metering and enterprise integration. One key and one bill, with quotas and budgets controlled per team.
Available
Model access
Access models through one integration layer instead of one contract per provider. The catalog is on the Models page.
- Access
- Unified API, OpenAI-compatible requests
Available
API and routing
Authentication, request routing and gateway policy, with routing rules, quotas and budgets set per team.
Available
Metering and billing
Token-level metering that supports usage-based agreements, billing and reconciliation.
Available
Business integration
Bring model capability into software, agents and business workflows. TD VIBE CODING shows one example in three steps.
Go to the MaaS platform
The platform lives at maas.xmax.com. This site covers the overview and account setup.
产品页：原 Enterprise AI 页升级为 Application Factory（L4）。旧三产品（Agent Studio／Knowledge Engine／Security Mesh）按 PRD D-10 不再作为独立产品，统一收进本页能力描述。
- Home
- Products
- Application Factory
Application Factory
Coding agents that turn a requirement into a deployable industry application, with people holding the quality gates. The goal is to turn one-off projects into a repeatable delivery process.
By project
Five stages
Requirements, architecture, coding, testing, deployment and operations. Each stage has its own agent, and all three layers below are used.
Every project
Human oversight
Requirements approval, code review, industry-expert validation and compliance review. Judgment calls stay with people.
In development
Reusable assets
Component library, industry packages, specification templates, quality framework and delivery SOPs, so the second project is faster than the first.
How we measure
Delivery metrics
Task completion rate, human intervention rate, component reuse, delivery cost per application, lead time and token revenue contribution.
Model
Revenue model
Project implementation fees, token usage for both build and run, and platform subscriptions with maintenance.
Start with an industry application
apMZoomAI is the first industry application. See the case study.
模型页：新增。现网无模型目录（商业面在登录后）。价格全部询价；上下文长度与版本按平台文档核对。
- Home
- Models
Model catalog
Open models served on XMAX capacity. Every model is available through the shared API and as a dedicated endpoint.
| Model | Developer | Good at | Context | Access |
|---|
Pricing is quoted per agreement. Model names belong to their developers. Versions and context lengths are confirmed in the platform documentation.
Need a model that is not listed?
We add open models on request for committed workloads.
解决方案页：现网 "Applications" 9 个独立页，每页三段长句 + 大量 "would not / does not claim"。改为一页 9 段：一句话、三个具体动作、用到哪几层。删除太空"在轨推理节点"、金融"量化交易"等无依据表述；互娱低清图问题（顾问第 5 条）随改版不再使用场景大图。
- Home
- Solutions
Solutions by industry
The same four-layer line serves very different work. Here is what teams in each industry use them for.
Have a different use case?
Tell us the workflow and we will map it to the right layer.
Trust Center：保留升级。现网 /trust、/security-governance、/sla 三页合并为一页分区；四步审查和 SLA 写法沿用。删除 "Named providers, traceable agreements" 空区块。
- Home
- Trust
Trust Center
How we decide who can use our GPUs, how we protect your data and what we commit to in production.
Company facts
Deployment review
Before any workload runs on our GPUs, we complete four checks.
Export controls and sanctions
Hardware destinations, end users, end uses and remote access are reviewed against applicable U.S. export controls and sanctions by qualified compliance staff. Customers provide accurate ownership, affiliation and use information as part of the review.
Security and data
Access
Role-based access, credential protection and review of every administrator.
Logging
Workload and administrative actions are logged and kept for audit.
Data handling
Retention and deletion are set per agreement. [Confirm: we do not train models on customer data]
Service levels
Availability targets, response times and service credits are set in each customer agreement. Planned maintenance is announced in advance. Incidents are triaged by severity with a named owner and a written review.
Due diligence requests
For security questionnaires, procurement reviews or compliance documents, write to info@xmax.com [security@ 待定].
公司页：现网 About 讲 "XMAX 集团 AI 能力承载主体"，同一句话全站出现 5 次；展示 10 个拟设立的 Pte. Ltd. 实体（顾问第 1 条）。改为一个主体故事 + 愿景/使命/目标 + 路线图 + 管理层；子公司表删除，已注册实体移至 Legal entities。
- Home
- Company
We build and run the full AI stack
XMax AI Inc. is the AI capability and industry service carrier of XMax Group, advancing platform capabilities, productized output and industry deployment. The parent, XMax Inc., is listed on Nasdaq under the symbol XMAX. We operate from Commerce, California and Singapore.
The group behind the platform
XMax Inc. is a Nasdaq-listed company with three legs: frontier technology investments, power semiconductors and the AI operating platform. XMax AI operates the third.
01
Investment portfolio
Frontier technology exposure through investment structures; details in the filings.
02
Power semiconductors
An acquisition agreement targeting advanced power semiconductor technology; details in the filings.
03
Operating platform
XMax AI: the four-layer production line, from NeoCloud to Application Factory.
Vision and mission
Vision
Every company should be able to run frontier-class AI on infrastructure it can trust and afford.
Mission
We build and operate the full AI stack, from GPU data halls to model APIs, so enterprises can put AI into production with one accountable partner.
Goals
- Capacity
- B300 in serviceTarget [Qx 2027]
- Platform
- Every model, two waysShared API and dedicated endpoint
- Reach
- [美国、日本、韩国、东南亚]Enterprise customers served
Roadmap
Timelines are forward-looking and may change.
Partners
Leadership
管理层：补职业照、3–4 句简介、全员 LinkedIn（顾问第 2 条：董事长和 CEO 缺链接）；删除现网显示的国籍。
Offices
Commerce, California
6565 E. Washington Blvd
Commerce, CA 90040, USA
Singapore
9 Battery Rd #15-108, MYP Centre
Singapore 049910
Investors
XMax Inc. filings and investor relations
Investor site [URL TBC]
新闻页：现网为 "Coming Soon"（新闻数组写死在代码里且为空）。改为 CMS 驱动 + 分类筛选 + 每条链接新闻稿或 SEC 文件。日期与标题一律按已发布原文。
- Home
- News
News
Announcements from XMAX AI and XMax Inc., with links to the original press releases and filings.
Media enquiries: info@xmax.com [press@ 待定]
联系页：删除页头背景图里重复的 "CONTACT US" 字样（顾问第 8 条）；表单补需求类型、规模、校验、发送中/成功状态与编号、同意条款与防机器人字段（顾问第 9 条），提交写入 Lark/CRM 并发确认邮件。
- Home
- Contact
Plan your capacity with us
Tell us what you want to run. A solutions engineer will reply within [1 business day] with options, timing and next steps.
Emailinfo@xmax.com
Security and procurement reviewsAsk for our due diligence pack in the form.
InvestorsSee XMax Inc. investor relations
法律实体页：新增。只列已注册实体（替代现网关于页的 10 个拟设子公司）。名单由法务提供。
- Home
- Legal entities
Legal entities
The companies that contract with customers and suppliers for XMAX AI.
| Entity | Jurisdiction | Role |
|---|---|---|
| XMax Inc. | United States | Parent company, Nasdaq: XMAX |
| XMax AI Inc. | [注册州], United States | Operating company for XMAX AI |
| [已注册实体，如有] | Singapore | [角色] |
新页面：Partners 不占一级导航（PRD D-03），从 Solutions、首页、页脚进入。已具名伙伴沿用分组与措辞，Logo 逐家许可，未授权不留空框。
- Home
- Partners
Build with us
Work with us at the infrastructure, inference, API or application layer, in four ways: GPU rental, reserved capacity, token consumption and application development.
Three partner types
Infrastructure
- GPU supply
- Data centers
- Connectivity
- Regional deployment
Technology
- Models
- Inference optimization
- AI services
- Integrations
Commercial
- Enterprise distribution
- API resale
- Industry solutions
Four ways to work together
GPU rentalCapacity by the hour or the term
Reserved capacityNodes and clusters locked on long terms
Token consumptionModel services billed on usage
Application developmentIndustry applications, delivered and run
Who we work with
Tell us how you want to work together
The partnerships team replies within a week.
新页面：据演示稿第 2、4 页。财务三数来自 10-Q（2026-06-30，未经审计），必须带日期与 unaudited；收入数字等 IR 放行；机构股东名单不出现（PRD 第七节第 2 条）；投资标的点名前需确认申报已覆盖。
- Home
- Investors
Investors
XMax Inc. (Nasdaq: XMAX) holds the AI operating platform together with frontier technology exposure. This page summarizes; the filings govern.
Three pillars
One long-term technology thesis, three legs: frontier technology investments, power semiconductors and the AI operating platform.
01
Investment portfolio
Frontier technology exposure through investment structures. [Names pending IR confirmation of filing coverage]
02
Power semiconductors
An acquisition agreement targeting advanced power semiconductor technology; see the filings.
03
Operating platform
XMax AI: NeoCloud, Token Factory, MaaS and industry applications.
Financial highlights
As of June 30, 2026, unaudited, source Form 10-Q.
$121.1M
Total assets
Jun 30, 2026, unaudited
$34.6M
Cash and equivalents
Jun 30, 2026, unaudited
$45.3M
Working capital
Jun 30, 2026, unaudited
Revenue
Recognized revenue from the AI business. Figures follow quarterly disclosure; anything not yet disclosed is not shown.
$1.10M
AI revenue, H1 2026
Pending IR wording
$4.5M
AI revenue, Q3 2026
Pending IR wording
Up to $25M
AI API agreement value
Framework ceiling, not recognized revenue
Revenue and agreement values follow IR-approved wording. The roughly $200M pipeline is opportunity and is not shown here.
Filings
Complete financial statements, risk factors and material events are governed by XMax Inc. filings with the U.S. Securities and Exchange Commission. Investor relations site [URL TBC]
新页面：据演示稿第 6 页。三段时间轴，全部挂"计划"与前瞩性声明；Application Factory 不给日期。
- Home
- Roadmap
Roadmap
Build capability, then scale delivery: platform releases, operational validation and a repeatable application model.
Planned and subject to change. See the forward-looking notice in the footer.
Two parallel tracks
Token Factory
Validate, then scale, then ecosystem. Prove the line on one site, then replicate across capacity and models.
Application Factory
Start with apMZoomAI as the first industry application, build reusable components, then replicate into other industries.
新页面：按 PRD v6.1 第 9 节。官网有两类非人类访客：抓取内容的 agent 与代表客户询价的 agent。此页说明可读什么、可做什么、限流与边界；入口只在页脚与 llms.txt，不占一级导航。
- Home
- For agents
For AI agents
This site serves both people and agents. Machine-readable endpoints, crawl policy and the inquiry API are listed here.
What you can read
| Endpoint | Contents | Format |
|---|---|---|
| /llms.txt | Site summary, key facts, endpoint index | text/plain |
| /llms-full.txt | Full site as plain text | text/plain |
| /facts.json | 14 facts, each with source and date | application/json |
| /pipeline.json | GPU pipeline: region, model, quantity, status, target date | application/json |
| /models.json | Models, developer, capabilities, context, access | application/json |
| /releases.json | Release calendar | application/json |
| /<path>.md | Markdown source of any content page | text/markdown |
All JSON carries schema_version and updated_at. Numeric fields must include source and date; figures not yet cleared return a placeholder marker, not a value.
Crawl policy
| Type | Examples | Policy |
|---|---|---|
| Search and citation | Googlebot, Bingbot | Allowed |
| AI assistants | ClaudeBot, GPTBot, PerplexityBot | Content pages allowed, /api/ disallowed |
| Training crawlers | CCBot, Google-Extended | [Legal review pending; content allowed by default] |
| Unrecognised | — | Content allowed, rate limited to 1 req/s |
How to request a quote
There is no machine inquiry API. Capacity and pricing are agreed per order and require a person, so quotes go through two human channels only: the contact form or email. Agents may read this page and the JSON endpoints to prepare a request, but submission happens through the form, not an API.
Contact form
Four request types: GPU capacity, model API, enterprise AI, partnership. A solutions engineer replies within two business days.
info@xmax.com. Include company, region, GPU count and start date for a faster reply.
No prices are returned and no capacity is committed automatically. Every inquiry enters the same human review and compliance process.
Need a person?
Sales and compliance questions go through the contact form.
We could not find that page
The link may be out of date. Pages from our previous website have moved.