Generative AI Enterprise
Posts
Goldman Sachs: How AI is re-platforming the economy

Goldman Sachs: How AI is re-platforming the economy

Plus, market share insights, Meta case study, and more.

Lewis Walker
March 12, 2025

WELCOME, EXECUTIVES AND PROFESSIONALS.

AI is re-platforming the economy, forcing existing technology stacks to evolve to fully harness AI’s potential. Goldman Sachs explores these profound shifts and the opportunities they are creating.

Since the previous edition, we've reviewed hundreds of the latest insights on best practices, case studies, and innovation. Here’s the top 1%...

In today’s edition:

Goldman Sachs explores AI re-platforming.
The leading generative AI companies.
AI transformation needs different leadership.
Meta: Uncovering soccer talent faster.
The future of AI research.
Transformation and technology in the news.
Career opportunities & events.

Read time: 4 minutes.

PLATFORM TRANSFORMATION

Goldman Sachs explores AI re-platforming

Image source: Goldman Sachs

Brief: Goldman Sachs published insights on AI’s role in re-platforming the economy, transforming how software and technology operates to drive value through three key vectors: compute, software, and data.

Breakdown:

Compute has progressed from traditional CPUs to powerful GPUs, with players like Nvidia, while newer CPUs are also emerging for specific AI workloads.
Software has evolved from application-centric to data-centric design, making data a foundational input for application development.
VP Ashraf stated: “These shifts have produced a '3D chessboard,' with companies from vastly different backgrounds competing for leadership.”
Major tech shifts typically see financial gains first captured by semiconductor and hardware companies, then “blossom” in the infrastructure software layer.
Infrastructure software (e.g. NVIDIA CUDA, Microsoft Fabric, Databricks) is central to AI re-platforming, linking hardware innovation at the bottom of the stack to application development at the top.
As in previous eras, infrastructure software will be the next investment frontier following outsized CapEx in AI hardware (e.g. GPUs).

Why it’s important: Enterprises who seize the platform shift early will lead in redefining existing markets and creating new ones. Goldman expects smaller AI companies to consolidate with larger players, enterprises to strengthen through M&A, and investors to increasingly focus on higher layers of the tech stack.

MARKET INSIGHT

The leading generative AI companies

Image source: IoT Analytics

Brief: IoT Analytics released a 263-page report on top gen AI firms, drawing from 530+ projects. The analysis covers three enterprise segments: data center GPUs, foundation models and model management platforms, and gen AI services.

Breakdown:

In just two years, the enterprise gen AI market across the three segments assessed has surged from $191M in 2022 to $25.6B in 2024.
The AI hardware market, specifically data center GPUs, experienced vast growth climbing from $17B in 2022 to $125B in 2024.
Data center GPU demand doubled in 2024. NVIDIA holds 92% market share, AMD grows from 3% to 4%, Huawei has 2%, followed by Cerebras and Groq.
Microsoft leads in “foundation models and platforms” with 39% market share in 2024, followed by AWS (19%), Google (15%), and OpenAI (9%).
This excludes the market for monetized consumer and enterprise generative AI apps, like ChatGPT.
The services market is more fragmented. Accenture leads with 7% share, followed by Deloitte at 3% and IBM at 2%.

Why it’s important: There is still a lot of movement to be had in the gen AI enterprise landscape. Microsoft CEO Satya Nadella recently warned the current AI hype may mirror the 1999 Dot-Com bubble. But ultimately, winners will emerge: those who consistently deliver value beyond hype and proof of concepts.

EXECUTIVE INSIGHT

AI transformation needs different leadership

Image source: L.E.K Consulting

Brief: Strategy consultants L.E.K. released an 11-page executive brief, informed by 150 C-level executives, finding that AI success relies on redefining the relationship between strategy, execution, the C-suite, and the broader organization.

Breakdown:

35% of surveyed enterprises show early AI success and optimism. These findings reveal different strategies in navigating the “AI Delta.”
Enterprises bridging this gap between successful and poor AI strategy and execution exhibit three key leadership attributes.
A strong vision. Enterprises with a clear AI strategy know what problems they want AI to solve and avoid endless use case exploration.
The CEO and CFO are visibly committed to the AI strategy and transformation, even if execution resides with a digital executive.
Increased collaboration among CTOs, CDOs, CIOs, and other executives. As one CTO put it, "It's a business norm that's really changing."
A chief data executive adds: "AI transformation is a decade-long trend, if not multi-decade, and you must stay the course."

Why it’s important: The AI opportunity is here, yet many businesses are struggling to transform. While no one has all the answers for this long-term shift, the most successful executives have embraced similar approaches, driving gains in performance, competitiveness, and creating new revenue streams.

CASE STUDY

Meta: Uncovering soccer talent faster

Image source: Meta

Brief: Meta shared an 8-page case study on how Sevilla FC used Llama to build Scout Advisor, a gen AI solution that transformed searching and analyzing over 300,000 scouting reports into a fast, accurate conversational experience.

Breakdown:

Sevilla’s in-house tools excelled at structured data but struggled with unstructured data, including valuable expert opinions.
They used Llama 3.1 70B Instruct and IBM's watsonx to create Scout Advisor, providing conversational search and player summaries.
Prompt enrichment ensures idiomatic user questions have enough context for a thorough semantic search of Sevilla FC’s 300,000+ scouting reports.
For instance, “show me talented wings” is enriched to note that a talented winger "takes on defenders with dribbling, creating space and penetrating opposition."
Instead of reviewing "45 reports to know what my scouts think," Sporting Directors get natural responses with player performance in seconds.
Plus, open-source Llama enables Sevilla to customize and optimize models without constraints from third-party APIs, securely within its environment.

Why it’s important: With 48% of enterprises planning to leverage open-source in their AI implementations in 2025, this case study highlights how prompt enrichment and semantic search, common generative AI use cases, can drive meaningful value extracting and summarizing unstructured data.

INNOVATION RESEARCH

The future of AI research

Image source: Association for the Advancement of Artificial Intelligence

Brief: The AAAI published a 91-page report on the future of AI research to help researchers in academia and industry, policymakers and broader stakeholders, navigate recent AI developments and challenges in driving innovation.

Breakdown:

The study was conducted by 25 AI researchers and informed by 475 survey respondents from academia and industry.
It highlights how AI researchers increasingly work in enterprise settings, where hardware and resources are more readily available.
It covers 17 AI research topics, exploring their history, trends, challenges, and perspectives from those involved in the study.
Topics include AI reasoning, trust, agents, evaluation, embodied AI, sustainability, perception vs. reality, hardware, AGI, geopolitics, and more.
For instance, the study notes the emergence of multi-agent systems in the late 1980s/90s, with interest growing this decade due to LLMs.
It reflects optimism for LLM-driven multi-agent systems but highlights the need to refine areas such as grounding and communication protocols before widespread adoption.

Why it’s important: With rapid change and competing narratives across enterprises and nations, it is important to be able to clearly identify the trajectory of AI research in a structured way. This helps collaboration on open challenges, driving innovation in AI’s capabilities and reliability for enterprise scenarios.

BCG shared a 15-slide report on Asia’s gen AI adoption, now second to North America. Informed by a survey of 240 CxOs and senior executives, it highlights traits of leading firms across countries such as China, India, Australia and New Zealand.

Microsoft published a 37-page report examining how agentic AI enhances business performance and growth. It includes the “three core agent actions”, and a blueprint for success, with a UK focus, though most insights apply globally.

Gartner’s 19-page report guides customer service leaders on future agent archetypes, examining AI's primary role as a decision-maker or teammate, and balancing cost with customer experience.

Sogeti, a Capgemini company, explores how Agentic AI applies the OODA (Observe, Orient, Decide, Act) loop, used by fighter pilots, for rapid, context-aware decision-making that anticipates and adapts to dynamic environments.

Deloitte shared five 20-40 minute presentations from the New York AI Summit, covering gen AI scaling strategies, trends, plus finance-specific insights. It also introduced CLAIRE® GPT, a gen AI-powered data management platform.

OpenAI launched new APIs for agentic apps, secured a five-year $11.9B AI infrastructure deal with CoreWeave, and is preparing AI agents priced $2K–$20K/month, targeting knowledge work and Ph.D.-level research.

ServiceNow is acquiring AI startup Moveworks for $2.85B, one of 2025’s largest AI acquisitions. The deal aims to integrate "ServiceNow’s agentic AI and automation strengths with Moveworks’ AI assistant and enterprise search technology."

Manus AI gained significant attention with impressive demos but isn’t likely "China’s next DeepSeek moment." It's built on existing models like Claude Sonnet, but does showcase what’s possible with the right tooling and integration.

Anthropic’s annualized revenue hit $1.4B this month, up 30% in the first two months of 2025, following Claude powering Manus and the launch of Claude Code and 3.7 Sonnet, reinforcing its strong growth and expanding AI capabilities.

Meta began testing its first in-house AI training chip to reduce reliance on Nvidia and control AI infrastructure costs, according to Reuters. It forecasts 2025 expenses of $114B-$119B, with up to $65B in capital expenditure for AI infrastructure.