In the era of AI, it is not at all a "human defeat"! The hottest positions have emerged, and the stock market is focused on two major themes.

The "Anthropic Storm" that severely hit software stocks is still fermenting in the global stock market, and the selling pressure is accelerating and spreading to wealth management, real estate consulting, and any other traditional industry that seems to be completely disrupted by AI. The pessimistic expectation that "AI disrupts everything" is impacting various sectors like dominoes falling, causing a series of declines in industries from software and SaaS to private equity, insurance, wealth management, real estate, property management, and even logistics, with investors accelerating the selling pressure on potential "losers". However, even as human society fully enters the AI era, there are still some positions that are becoming increasingly hot, such as FDEs that embed models into enterprise workflows, with demand expected to grow 42 times between 2023 and 2025. The final pricing of the stock market may not be a simple "software loses, AI wins" scenario. The real favorable directions may be focused on two main themes: AI deployment and governance infrastructure, such as platform integrators that help companies effectively implement AI, and AI data center computing power and electricity supply chain. Over the past month, Anthropic, a leader in AI applications often referred to as the rival of OpenAI, has continuously released heavyweight products, triggering massive selling waves in the global stock market, particularly impacting enterprise software stocks, with major software giants like Microsoft, SAP, and ServiceNow still struggling to recover from the stock price declines since the "Anthropic Storm" in early February. The core factor behind this wave of software stock declines and the rotation of declines across various sectors is the pessimistic notion of "AI disrupts everything". Since February, this sentiment has swept through global financial markets, leading to a significant sell-off in the SaaS subscription software sector and the broader software sector. As fears grow about the proliferation of AI agent workflows like Claude and OpenClaw, the market believes that industries relying on seat subscription revenue models or labor-intensive business models, like insurance, real estate, trucking, and any other industry that may be disrupted by AI, will be completely overturned. The selling pressure led by the "Anthropic Storm" intensified last week with Anthropic releasing Claude Code Security - an AI-driven network security vulnerability scanner. This tool caused network security companies like CrowdStrike, Cloudflare, and Okta to plummet by 8% to 10% in a single trading day. Subsequently this week, after Anthropic announced that its Claude Code tool could help companies achieve low-threshold automation processes for traditional programming languages running on IBM systems, IBM experienced its most severe one-day stock price decline in more than 25 years. The value of human bridges cannot be ignored! Perhaps the most popular job in the AI era - FDE Some market anxiety is justified. The software economy has long relied on seat-based licensing models and hourly billing. If highly efficient AI agent workflows can autonomously perform tasks that were traditionally done by junior programmers or legal assistants, existing software companies may face pressure to defend their core revenue models. However, more turbulence reflects a broader market unease - investors have been discussing this for months - a structural debate around whether AI is enhancing or shrinking enterprise software. The reality is more complex. Anthropic itself points out that its legal tools do not provide final legal advice, and its security scanner still requires human approval for patching. In order for these AI agents to truly function in enterprise workflows, the most advanced AI labs globally are actively hiring a specialized role to bridge the talent gap at the technical level. One open secret in the enterprise AI wave is the importance of hiring a hybrid role known as Forward Deployed Engineers (FDEs). Initially popularized by Palantir - embedding engineers within government and military clients - the FDE model has now become a popular market expansion strategy in the AI era. According to a recent report by LinkedIn tracking global AI job positions between 2023 and 2025, demand for FDE and similar AI engineering roles has grown 42 times. While there have only been about 9,000 new FDE positions globally, they target the industry's biggest bottleneck: ensuring that AI can operate effectively in the real world. These individuals are not just pre-sales engineers or specialized IT support. FDEs are versatile engineers who work on-site with clients, navigate internal politics, and write production-level code to make models truly yield results. After two years of enterprise AI advancement, both OpenAI and Anthropic have reached the same conclusion: the unprecedented wave of AI infrastructure requires the FDE cohort to deliver on its promises. OpenAI sees them as a "parachute consulting force," with up to 50% of their job list and time spent traveling and embedding directly with strategic clients. Meanwhile, Anthropic focuses heavily on complex deployments in regulated industries like finance and healthcare, requiring a significant hiring of application-focused AI engineers like FDEs to build customized workflows. Large-scale cloud computing companies like Google and Amazon are also combining their cloud platforms with IT tech teams closely aligned with enterprise clients to capture vast AI workloads and expand their dominance. To attract this scarce talent - engineers who can debug an agent one minute and report to Fortune 500 executives the next - both AI labs, OpenAI and Anthropic, are offering top-tier compensation packages. OpenAI's base salary can reach up to $325,000, while related positions at Anthropic can go up to $400,000, with stock compensation easily exceeding $500,000. Denise Dresser, Chief Revenue Officer at OpenAI, says that these talents are "seasoned engineering professionals in the traditional sense, accustomed to dealing with the complex deployment of large AI models, working on an enterprise scale, and having a good contextual understanding of business knowledge and technology." She emphasizes the importance of communication skills, as FDEs act as a bridge between product teams and enterprise clients. While adapting models to specific workflows, FDEs provide feedback to AI labs for model and product development. In one case, after embedding AI into a workflow, client managers were able to save about 90% of their time - the kind of quantifiable positive change that enterprise clients are seeking. But despite rising salaries and increasing demand, there remains an undecided question: how long will the FDE boom last? Dresser believes that this model is transitional. The goal is not to permanently do the work for clients, but to guide them in the early, complex stages of adopting AI until they can be self-sufficient. Currently, the AI arms race is not just about building stronger AI models. It is also about building human bridges that make these models truly usable. Whether this bridge becomes a permanent layer in the AI economy or fades away as tools mature will largely inform us about the duration of this enterprise-level AI supercycle. According to US employment data, the more an industry embraces AI, the more apparent the reduction in workforce seems to be. The chart that plots adoption rates against changes in unemployment since 2022 shows this unsettling correlation. The information industry (telecoms, media, publishing) has an AI adoption rate of around 70%, resulting in a 75% increase in unemployment. Other AI-intensive sectors, such as professional services and finance, also fall above the trend line. Research from the Dallas Fed this week found that in the top 10% of industries most exposed to AI, employment has declined by 1%, while total US employment has increased by about 2.5% since the end of 2022. Wage data may provide some optimistic reasons. During the same period, professions that rely heavily on implicit knowledge, judgment, and experience have seen continued wage increases. In the top 10% of industries most exposed to AI, wages have grown by 8.5%, higher than the average wage growth of 7.5%. The pattern that emerges is a widening gap: more positions are disappearing in AI-intensive industries, but workers with skills that AI cannot replace are still receiving substantial pay raises. The winners and losers forces being priced in by the stock market The AI era is not a complete collapse for humanity, but a comprehensive reassessment of positions and investment themes: the most popular job types are booming, while the logic of pricing technology stocks is being completely rewritten. Just as the prospects for AI-era career development like FDEs are very optimistic for "top-tier multifaceted technical talent," it is not unilaterally positive for the overall labor market. The unstoppable wave of AI is not a 100% positive catalyst or completely negative factor for the stock market. The first level of impact on the stock market is the revaluation of enterprise software and professional services, as seen when Anthropic launched a legal plugin that led to significant declines in companies like Thomson Reuters and RELX that provide single-function consulting software. More broadly, a Breakingviews study indicates that AI is challenging the "seat-based licensing + hourly billing" model of the old software economy. The BVP Nasdaq Emerging Cloud Index has significantly fallen from its peak, and revenue growth for some PE-supported software companies has decreased from 27% in 2021 to 10% in 2023. This implies that the clearest losers are those highly reliant on seat-based models, junior labor, templated workflows, and information asymmetry charges - including some vertical SaaS providers, legal/compliance information services, traditional code maintenance and low-value IT outsourcing, and some point tool-based cybersecurity companies. The ultimate pricing of the stock market may not be as simple as "software losing, AI winning". The true beneficiaries are likely to be two main categories: AI deployment and governance infrastructure, including cloud and model platforms, data governance, identity access, auditing/security, observability, workflow orchestration, and comprehensive integrated platform services that truly help companies operationalize AI. When AI significantly improves optimization efficiency and reduces marginal decision costs, platform software companies like Microsoft, ServiceNow, and SAP may see a positive feedback of "throughput enhancement + stronger unit economics" rather than being completely replaced. The second category is the most competitive global market logic around AI data center computing power and electricity supply chain, such as leading chip manufacturers, HBM/high-performance server/storage chip makers, high-performance AI server contract manufacturers, and core hardware equipment assemblers for AI data centers. This is mainly because once AI applications enter real production environments, the demand for massive inference loads, storage, networking, cooling, and power will continue to rise. The White House is currently pushing large tech companies to bear additional power costs for AI data centers, as the rampant expansion of AI infrastructure has put immense pressure on power grids and electricity prices. In other words, the core beneficiaries of the stock market in the AI era will gradually shift from "who has the model" to "who can smoothly, compliantly, and cost-effectively deliver and sustain AI large models into enterprises". Therefore, the core sectors benefiting from the AI era can be summarized as cloud software platforms and inference infrastructure, data center power/cooling/networking, AI governance and security, and high-end integrated and consultancy software platform services. The biggest losers are those who essentially digitize human processes but fail to form data moats and system-platform closed loops in the old SaaS software model. Overall, the AI era in the stock market will transition from "who has the model" to "who can deliver and run AI large models in enterprises steadily, compliantly, and cost-effectively". As a result, the core beneficiaries in the AI era are summarized as cloud software platforms and inference infrastructure, data center power/cooling/networking, AI governance and security, as well as high-end integration and consultancy software platform services. The biggest losers are the companies whose products are essentially "digitizing human processes but do not form data moats and system-platform closed loops" in the old SaaS software model.