Anthropic, Meta 독점 서비스용 영업팀 모집, 1인당 최대 320만 달러

On a quiet note, Anthropic recently posted a mysterious job opening: "Mega Account Executive, Meta"—in plain terms, a sales position dedicated to Meta as a super-tier customer. The annual salary ranges from $380,000 to $450,000, equivalent to approximately 3.2 million RMB.

And this sales role serves only one customer: Meta.

What? Annual income of over 3 million RMB by servicing just one customer? It sounds appealing...

But the position vanished quickly. Launched on September 3rd and closed on September 9th, it existed for only one week. The recruitment page now shows that the position has stopped accepting applications.

What happened—was it filled so quickly?

**What does this 3.2-million-RMB "To M" sales role actually entail?**

The job posting outlined specific requirements: work location in San Francisco or New York; minimum 10 years of enterprise sales experience; responsibility for the complete sales cycle from opportunity discovery and technical evaluation through procurement and contract signing; ability to penetrate multiple business divisions within Meta; relationships spanning from frontline employees to C-suite executives; and relaying Meta's feedback to Anthropic to influence product roadmap.

(C-suite refers to the highest corporate management level—CEO, CFO, CTO, CPO, COO, etc.)

This million-dollar salary isn't without its challenges. Anthropic requires candidates with at least 10 years of enterprise sales experience, familiarity with procurement processes at large tech companies, and ability to handle complex deals with long cycles and multiple stakeholders.

Upon onboarding, the salesperson must identify business units within Meta that could potentially use Claude, from uncovering needs and organizing technical testing through procurement and contract signing. Relationships cannot be limited to the procurement department—the salesperson needs to establish multilayered connections from frontline engineers and technical teams all the way to top executives.

In short: you need to know what AI tools Meta engineers are currently using, while also convincing management to continue paying for Claude. Additionally, you must bring Meta's feedback back to Anthropic to help the product and engineering teams adjust their roadmap. When new use cases emerge, mobilize Anthropic's internal product, marketing, and technical resources to expand the business further.

The job description emphasizes: "identify new use cases and business divisions within strategic customers to drive business expansion." This means embedding yourself long-term within Meta, finding footholds for Claude across different teams, and scaling scattered usage into larger enterprise contracts.

The requirements genuinely rival the recently popular FDE (Field Deployment Engineer) role.

Moreover, the $380,000–$450,000 annual salary operates on an OTE mechanism. OTE stands for "On-Target Earnings," the total annual income an employee is expected to earn upon meeting set performance targets. It typically consists of base salary plus variable compensation—sales commissions or bonuses. So the $450,000 figure isn't guaranteed upon hire; actual income depends on hitting the company's targets.

**Meta is building models but aggressively buying Claude**

So the question arises: Meta is developing Llama and Muse, so how did it become an Anthropic power customer?

Primarily because Claude Code has already deeply penetrated Meta internally. Earlier this year, Meta engineers were already using Claude Code extensively. According to The New York Times, Meta's potential annual spending on Anthropic could reach $10 billion.

According to Reuters, as of late July, Anthropic's annualized revenue had exceeded $6.5 billion. Meta's potential spending ceiling alone represents approximately 15% of that figure.

Moreover, enterprise large customers are currently critical for Anthropic's revenue. In February, Anthropic disclosed that the number of customers with annual spending exceeding $100,000 grew 7-fold over a year; customers spending over $1 million annually increased from 12 two years ago to over 500. Eight of the Fortune 10 have become Claude customers.

Claude Code has been a major driver of this growth. Since officially launching in May 2025, it reached $1 billion in annualized revenue in just half a year; by February 2026, that figure exceeded $2.5 billion, approximately 18 billion RMB—more than doubling from year start. By early 2026, Claude Code's enterprise subscription count had grown 4-fold, with enterprise customers contributing over half its revenue.

It's no wonder Anthropic established a dedicated million-dollar sales position for Meta—this is simply the extreme version of their enterprise customer strategy. A company with tens of thousands of employees and multiple engineering and product teams means each new use case adds potentially significant model API call volumes.

Even Meta's consumer-facing Agent project shows traces of Claude. According to The Information, Meta's internally codename "Hatch" consumer-grade Agent initially used Anthropic models for development and testing. The product later launched as Muse, with the public version running on Meta's in-house Llama foundation.

This illustrates their relationship well: Meta can use Claude to validate products and accelerate development, then shift the underlying capabilities to its internal stack once its own models mature.

However, Meta is also looking for ways to reduce these expenses. Reports indicate Meta has built an internal management platform tracking in real-time the costs employees incur using external AI tools, with budgets set per department. Simultaneously, the company is promoting its internal engineering tool MetaCode and externally launched Muse Code, hoping to reduce engineering teams' dependence on Claude Code and similar products.

**One More Thing**

The relationship between Meta and Anthropic has at least three dimensions:

**1. Model competition**: The two companies compete directly on model capabilities. Meta is pushing Llama and Muse Code to catch up to Claude and reduce internal reliance on Claude Code.

**2. Model purchasing relationship**: Meta is also an important Anthropic customer, buying its services for internal R&D, programming collaboration, and product validation.

**3. Data center compute procurement relationship**: At the infrastructure level, money might flow in reverse. Around the same period, Anthropic discussed with Meta the purchase of data center compute capacity, potentially spending up to $10 billion over two years. However, like Meta's model spending, this compute procurement hasn't become a publicly confirmed transaction between the two.

One $10 billion flows from Meta to Anthropic for models; another $10 billion potentially flows from Anthropic to Meta for compute. Competing on models in public, ordering from each other behind closed doors—Silicon Valley is indeed complex.

It appears that in the AI wars, even super customers need "one-to-one service." Anthropic's website currently lists 112 sales-related positions, mostly divided by industry, region, or customer type—automotive, public sector, startups, AI-native companies. By contrast, a mega customer like Meta has its own designation directly written into the job title, preceded by "Mega Account."

This reveals a shift in large language model commercialization: the next wave of revenue growth for leading model companies increasingly depends on a handful of super customers capable of deploying models across tens of thousands of employees and multiple business divisions. Such orders can't be secured from the price list on the API website. Someone needs to embed within the customer, drive engineers to test, negotiate with procurement, explain ROI to executives, and funnel frontline needs back to the product team.

This may explain why Field Deployment Engineers have become increasingly popular domestically and internationally recently.