The 2024 MAD Landscape: An overview by AIport

A short version of the annual landscape from FirstMark

The 2024 MAD Landscape has just come out, investigating the state of affairs in the fields of data, machine learning, and AI. Since we at AIport love this type of research — having published the Global GenAI Landscape of our own just a couple of weeks earlier — we thought it would be a good idea to give you our succinct take on what the MAD Landscape is all about.

Who and what:

• This is the tenth “state of the union” publication from Matt Turck and his colleagues at FirstMark (a VC firm from New York), as well as other collaborators, including researchers from Go Fractional.  

• The whole publication is divided into three parts: Part 1 – the actual landscape (here’s the interactive version), Part 2 – the 24 “hot” themes of 2024, and Part 3 – financing (including mergers and acquisitions, as well as IPOs).

Part 1: The Landscape

• There are 2011 companies featured in the MAD Landscape this year, with 578 new entrants compared to last year.

• These numbers are explained by the still ongoing data infrastructure cycle that began roughly a decade ago, as well as the more recent GenAI wave that turned into a tsunami with the launch of ChatGPT.

• Three new categories were added next to MLOps: AI Observability, AI Developer Platforms, and AI Safety & Security.

• Conversational AI was added to the Search category in Applications to reflect the current situation in the field (e.g., Bing is now integrated within Copilot, Brave has incorporated GenAI, there’s Perplexity, etc).

• A new category of Local AI was created to represent a trend of providing infrastructure tooling for regional development. 

• A necessary distinction is made between Commercial AI Research (e.g., Sam Altman’s OpenAI) and Nonprofit AI Research (e.g., Elon Musk’s xAI).

Part 2: The themes

The authors of the MAD Landscape outline the following noteworthy topics relevant to AI and data development in 2024:

• Structured vs. unstructured data: This point highlights the evolving tech landscape, where unstructured data’s role in GenAI is gaining prominence over structured data’s traditional analytics applications.

• The Modern Data Stack (MDS): Once a pinnacle of tech innovation (2019-2021), it’s now under pressure due to its high integration costs and the shift in focus towards GenAI.

• Data infrastructure: We’re witnessing a pivot towards AI in the data field, which is approaching market consolidation, and also seeing startup challenges (shorter runways), as well as functional expansions by leading firms. 

• Databricks vs. Snowflake (key data players): Snowflake faces growth slowdown and product pressure, lagging in AI integration, while Databricks emerges as a GenAI leader, with Microsoft’s new Fabric platform a challenger to both.

• Business intelligence: Despite slow transformation and ongoing dominance by established solutions, GenAI is poised to revolutionize analytics and disrupt the industry by enhancing data extraction and refining insights.

• The modern AI stack: The rise of GenAI is leading to the evolution of scale-ups in areas like vector databases and frameworks, mirroring the once-dominant Modern Data Stack.

• The latest AI hype cycle: We’re currently in the third intense AI hype cycle since GenAI’s surge in November 2022, driven by broad audience appeal and significant VC interest. 

• Here to stay or a fad?: Amidst widespread adoption of GenAI, concerns arise regarding the authenticity of its impact vs. experimental enthusiasm, particularly in enterprise spending and long-term consumer engagement.

• LLM companies: As leading providers thrive, adopting a full-stack approach akin to cloud vendors, and open-source models emerge as strong competitors, questions linger over potential commoditization.

• Hybrid future: The emergence of Small Language Models (SLMs) alongside LLMs like GPT, Claude, and Grok reflects a move to hybrid architectures in enterprise deployments and specialized models for diverse use cases.

• TradAI vs. GenAI: The launch of ChatGPT highlights a new direction among developers, where Traditional and Generative AI are combined for specific applications, often complementing each other.

• From thin to thick: Startups evolve from using basic “thin wrappers” to crafting “thick” full-stack solutions, focusing on niche markets and deep AI customization, while avoiding the “kill zone” of Big Tech.

• Novel areas: Edge AI promises local, device-based intelligence without GPUs, while AI agents aim to revolutionize automation with “text to action” capabilities (though still lacking in system memory and predictability).

• AGI or a plateau?: GenAI’s path to Artificial General Intelligence (AGI) is uncertain, facing resource limits (compute and data) as well as “reasoning” dead-ends, despite methodologies like AlphaGeometry and new GPUs like Blackwell. 

• The GPU wars: Amid NVIDIA’s soaring valuation and market dominance for AI-ready GPUs, emerging competition and innovations are challenging its leading position, while the chip industry faces potential overproduction risks.

• Open-source AI: As freely available AI models flood the market, often with positive effects, the space also must address the issues of oversaturation and the viability of so many “weekend” projects.

• AI economics: While the end-user’s GenAI costs are being driven to the bottom by vendors in the midst of competition, providers themselves face high building and maintenance expenses, calling profitability into question.

• AI politics: The quote below from Matt Turck says it all. 

• OpenAI: The company’s $86-billion valuation is impressive, but the pace might not be sustainable, given OpenAI’s complex relationship with Microsoft, which raises concerns about whether this partnership will eventually break up.

• 2024: After 2023’s frenzied AI adoption and consulting boom, 2024 holds promise for GenAI’s significant enterprise impact, navigating issues like tool selection and deployment, model hallucinations, and skills shortages.

• SaaS: The ongoing debate questions whether AI will render Software-as-a-Service obsolete, suggesting instead a future where SaaS evolves alongside AI, with the latter enhancing rather than replacing traditional software solutions.

• AI and VC: The role of investment is being examined — from needing huge sums to develop fresh AI models that most VCs can no longer afford to having new ultra-efficient, AI-powered firms that can bypass VC funding paths entirely. 

• Consumer revival: GenAI is changing consumer technology by challenging Google’s search monopoly (e.g., You.com and Metaphor), introducing AI companions, and creating hyper-personalized entertainment.

• AI and blockchain: While there’s no MAD crypto section, the authors outline how AI’s alliance with blockchain promises to challenge AI’s centralization but also poses the perennial crypto challenges of scams and overhype.

Part 3: Financing

• In terms of funding, the data sector saw little activity in late 2023 and early 2024, with Databricks (raised $684 million, $43.2 billion valuation) and Sigma Computing (raised $340 million, $1.1 billion valuation) being rare exceptions.

• Within the AI sector, there’s been a large concentration of capital in the hands of a select few startups, including OpenAI (raised $10.3 billion, $86 billion valuation); Anthropic (raised $6.45 billion, $18.4 billion valuation); Hugging Face (raised $235 million, $4.5 billion valuation); and Inflection AI (raised $1.3 billion, $4 billion valuation). 

• Microsoft, Google, and NVIDIA were by far the most active AI investors in 2023.

• The mergers and acquisitions (M&D) market has remained quiet since last year. Among the most notable data- and AI-related M&D deals were Broadcom (a semiconductor manufacturer) acquiring VMWare (a cloud computing company) for $69 billion; Cisco (a networking infrastructure company) acquiring Splunk (a monitoring platform) for $28 billion; and Microsoft orchestrating the “non-acquisition acquisition” of Inflection AI.

• The “Magnificent Seven” stocks (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla) grew almost 50% last year.

• Among noteworthy IPOs, Reddit went public in March with a valuation of $6.4 billion, following the move to license its content for AI training. Likewise, Astera Labs, a semiconductor firm known for providing intelligent connectivity to AI players, also went public, achieving a $5.5 billion valuation.

The bottom line

The 2024 MAD Landscape is a comprehensive piece that offers invaluable insights for AI enthusiasts, data scientists, current or future investors, and tech aficionados alike.

If we were asked to generalize across all observations and themes, the landscape’s leitmotif would be that GenAI adoption is rapidly taking hold — not only through LLMs but now also through SLMs — which is, in turn, leading to a redefined field of data science, with a move away from structured data and MDS.  

While the research carried out by Matt Turck and his colleagues is undeniably top-notch, the landscape’s main drawback (apart from its intimidating size) is that its primarily Western perspective overlooks significant AI contributions from other regions. Many if not most AI players from Asia are absent from the landscape altogether, including big names like Huawei, Xiaomi, and Naver. Additionally, smaller Western players (e.g., AI search engines like Komo and Andi) aren’t covered in the report either.

Having said that, the landscape is of the highest quality, and though somewhat limited in its scope, it rewards the patient reader with exclusive content and critical findings within the AI sector.