Contents
- What Ramp’s enterprise AI spending data says about the OpenAI-Anthropic race
- New signals among paying customers — and what they mean for how marketing teams choose models
- Technical constraints are changing how companies spend on advertising and operations AI
- What OpenAI and Anthropic mean for Vietnam: how businesses should read the signal
- What to do so your advertising AI choice is not swept up by brand effect
- References
Enterprise payment data is showing that the AI race is not just about how prominent a model is, but about how well it plugs into real procurement workflows. For Vietnamese marketers, this is an important signal because AI budgets in advertising and operations usually follow usability, control, and explainability—not just technology reputation.
When one side speeds up while the other still holds an advantage, the market is making one thing very clear: businesses do not “choose once and be done,” but shift with each model launch, price change, data condition, and level of fit for the job. The question is therefore not who wins outright, but what mechanism causes spending to move.
- Key point:
- Ramp only reflects more than 70,000 U.S. businesses using its payment suite, but it is enough to show the direction of AI spending in the enterprise segment.
- OpenAI has regained growth momentum in Q3 to date, even though Anthropic still holds a higher share among Ramp’s paying customers.
- The enterprise AI market is expanding, so the fight for share is happening while the total number of companies paying for AI keeps rising.
- The deciding factors are not only whether a model is strong or weak, but also price, data rules, and how easily it can be put to work in practice.
What Ramp’s enterprise AI spending data says about the OpenAI-Anthropic race
Ramp is a spend management and corporate card company. Its data covers more than 70,000 U.S. businesses, with spending reaching billions of USD through its bill pay and corporate card systems. This is not the whole market, because many large enterprises use American Express or other providers, but it is still a slice close enough to real buying behavior to read the trend.
In that slice, OpenAI was once ahead in both enterprise customers and mainstream users, but it lost the number-one position among Ramp’s paying customers in May. By July, Anthropic still held the larger share. Even so, the latest data shows OpenAI growing faster in Q3 to date. That means the market has not locked into a single choice.
What matters for marketers is that this shift is not driven by brand communication alone. It follows the way businesses test, switch, and then return when a new model does the job better. For AI tools used in advertising, content, analytics, or developer support, spending is usually decided very pragmatically: can it plug into the workflow, can data be controlled, and is it worth the money relative to the output?
New signals among paying customers — and what they mean for how marketing teams choose models
Recent conditions suggest OpenAI has been lifted by several new models with strong appeal among developers, while Anthropic has faced pressure because some usage terms have made enterprise users think more carefully. This is a signal about buying behavior, not a declaration of victory.
OpenAI and Q3 growth: what marketers need to understand about “faster”
Ramp’s data shows OpenAI growing faster than Anthropic in Q3 to date. The source of this signal is enterprise customer usage on the Ramp platform, not the global revenue of either company. The lesson here is that growth speed in a large enough payment channel can reflect real demand, but it does not mean the whole market has been won.

For marketing teams, the key reading is this: whichever model creates an early deployment advantage will pull trial and expansion spending with it. When a new model performs better at writing content, supporting code, or integrating into internal tools, budgets tend to shift quickly. But because AI purchasing is still flexible, that advantage will not last if the vendor cannot maintain quality and real-world usability.
Anthropic and data-retention rules: when a technical barrier becomes a buying barrier
Anthropic drew criticism after warning Fable users that they had to allow 30-day data retention. According to TechCrunch, Fable is Anthropic’s higher-end model tier. This is a clear example of how data terms are no longer a legal appendix, but a factor that directly affects purchase decisions.

In marketing, especially in advertising, CRM, and behavioral analytics, input data always touches privacy and internal control processes. If an AI tool forces a team to accept a retention level that does not fit, the real cost is no longer just the model fee, but the level of operational risk. That is why some tools can slow down even when their technical quality is still strong.
An expanding market: the share battle is happening while the number of paying companies keeps rising
Ramp says the share of businesses in its customer base paying for AI passed 50% in March and reached nearly 56% in July. That figure shows the market has not yet reached saturation. The two companies can keep taking share from each other, but the overall pie is still growing.

For marketers, this is why the race should not be read as a simple “pick a side” game. When the market is still open, businesses can test multiple models for different tasks: one model for content creation, one for analytics, one for internal operations. What matters is standardizing the selection criteria: output quality, speed, integration ability, data, and actual cost.
Technical constraints are changing how companies spend on advertising and operations AI
The core point in Ramp’s data is not who is slightly ahead, but that businesses are willing to switch vendors when technical constraints change. For advertising and marketing, AI is only bought at scale when it can enter real systems: dashboards, content workflows, internal tools, approval processes, and measurement.
Workflow integration: what decides which model gets bought next
A model can be excellent, but if it is hard to plug into the process, spending will still be difficult to scale. Ramp’s data shows businesses do not remain absolutely loyal to one vendor. They move according to their sense of implementation effectiveness, especially among technical users and teams that need fast output.
Vietnamese marketing teams usually buy AI for two goals: reducing work time and improving output quality. If a tool cannot connect to existing work, the team has to add another manual layer, and the total cost of ownership rises. In that case, a model that looks stronger on paper may still lose to one that is easier to use.
Price, data, and trust: the trio that changes buying decisions
The issue here is not just list price. It also includes data-retention terms, the ability to explain decisions to legal teams, and the level of trust from operations teams. Ramp shows that businesses are willing to switch when one side offers a more attractive new model, but they are also willing to return if the other side performs better in actual use.

For advertising, this is especially important because customer data, campaign materials, and measurement information are all sensitive. A tool may be very strong on productivity, but if it makes internal control teams uneasy, its use will be limited. At that point, the buying decision no longer sits only with the marketing department, but passes through several layers of approval.
What OpenAI and Anthropic mean for Vietnam: how businesses should read the signal
The Vietnamese market usually lags the U.S. by a step, but buying behavior is quite similar: prioritize what can be used immediately, what can control data, and what will not break existing workflows. That is why the OpenAI-Anthropic race is a useful indicator of how Vietnamese businesses may choose AI tools for advertising, content, and internal automation.

For performance, content, and brand teams, the lesson is not to pick one side early and lock the budget. The lesson is to design an internal evaluation framework: which model is for speed, which is for sensitive content, which is for analytics, and when to stop because the data terms are not suitable. Without that framework, buying decisions will be driven by the feeling that a product is famous.
In Vietnam, many businesses are also moving from experimentation to real use in tasks such as writing ad copy, summarizing sales materials, supporting customer service, and analyzing insights. The OpenAI-Anthropic race is a reminder that “real use” always brings governance questions with it. What you are buying is not just a model, but the way it fits into your operating system.
What to do so your advertising AI choice is not swept up by brand effect
- Evaluate models on four things: output, workflow integration, data terms, and total real-world cost.
- Do not choose a tool just because it is being talked about; test it on a real task such as ad writing, insight summarization, or analysis support.
- Define clearly which data can go into the model and which data must stay internal before rolling it out to the whole team.
- Set up a regular comparison process between models, because this market changes very fast; what is good today may not keep its advantage next quarter.
For Vietnamese marketers, the most important thing is to learn how to read enterprise spending signals as an operational indicator, not as a reputation race. When U.S. businesses can still switch back and forth between two major vendors, then in Vietnam the choice should be based even more firmly on what can be used, controlled, and proven effective.
See more marketing analysis and guides at https://marketing365.vn.
Follow more analysis from Marketing365 to stay updated on the latest marketing trends.
Read more articles in the same category at Digital Trends.



