Key takeaways

• Regulations
  • While the finalization of the EU AI Act is still pending, China implemented GenAI regulations in August. These new laws govern both AI content service providers and model providers and include measures against illegal and discriminatory content and the use of personal data.       
  • In the Western regions, there were some advancements, with a few guidelines and principles reported during the quarter (such as guidelines for US government employees using genAI and the Canadian government introducing a voluntary code of conduct); however, solid laws and regulations are yet to be materialized.

• GenAI Applications
  • We saw prominent disruptors such as OpenAI and Anthropic focusing on developing more advanced chatbots. These new chatbots can do more than just chat; they can handle voice interactions, generate images, and even write computer code. This shift represents a significant advancement in the field, promising more powerful and versatile AI tools in the future.
  • Funding in the space saw a decline (Q3: ~USD 270 million, down 56% QoQ). This could be a sign of diminishing enthusiasm within the industry for the smaller players, due to prominent tech giants' substantial influence.
  • A notable trend was more established disruptors strategically collaborating with incumbents to expand the accessibility of their solutions. Examples include Jasper’s partnerships with Google Workspace, Webflow, Zapier, and Make and Typeface’s partnership with Google Cloud and GrowthLoop to launch a unified marketing solution.

• Foundation Models 
  • We saw players in this space, disruptors and incumbents alike, launching new models (OpenAI’s DALL-E 3, StabilityAI’s FreeWilly, FreeWilly2, Diffusion XL 1.0, StableCode, and many more) with various capabilities, including answering complex questions in specialized domains, advanced text-to-image generation, coding, and music generation.
  • Tech giants continue to invest, as demonstrated by Amazon's substantial USD 1.25 billion investment in Anthropic, as it looks to strengthen its position in the space. Five players raised a total of ~USD 2 billion across six rounds in disclosed funding in Q3 2023 (~22% less than the previous quarter).
  • Companies have been actively forming strategic partnerships to broaden the accessibility of their models across various platforms, including 1) Meta’s partnerships with Microsoft, Alibaba, IBM, and Amazon to offer access to its Llama 2 family of models on these platforms, and 2) AI21 Labs partnering with Google to make its large language models (LLMs) available on Google Cloud Marketplace.

• GenAI Infrastructure
  • As the GenAI landscape evolves, companies are proactively seeking opportunities to optimize model efficiency and processing (e.g., VSORA Introducing Jotunn, an all-in-one single-chip family designed to enhance performance for GenAI inferencing processing). They are increasingly turning to hardware innovations designed to support edge processing that could facilitate reduced power consumption, enhanced processing speeds, and a heightened focus on secure data management.
  • Funding picked up in Q3 2023, with 16 disruptors raising a total of ~USD 913 million, 5x more than Q2. Databricks’ USD 504 million series I round was the highest funding amount recorded by a disruptor during the quarter.
  • Furthermore, as leading players in the cloud infrastructure domain continue to expand their capabilities, businesses are increasingly recognizing the value of partnering with them to create GenAI solutions.

Regulations: China moves ahead with regulations; developments yet to materialize in the West

Analyst take: The Chinese government has been proactively encouraging the advancement and acceptance of AI, implementing new rules to govern GenAI. This marks the country's efforts to strengthen oversight of this growing and changing technology. While strengthening the regulatory background, China also seems to be advancing the development of the technology to compete with its US and other counterparts; the move to approve the first batch of GenAI services to be publicly rolled out also marks a milestone for China's AI industry and its potential leaders, enabling them to compete with global tech giants. While several advancements including guidelines and principles were reported during the quarter in the West, the establishment of solid laws and regulations is yet to be seen. 

• In July 2023, the Cyberspace Administration of China, in collaboration with six other regulatory bodies, introduced new rules to govern GenAI, which came into effect on August 15. The new rules, which cover AI content services, including text, pictures, audio, and video, provided to the Chinese public, require GenAI service providers to obtain a license, monitor and prevent illegal and discriminatory content, conduct security assessments, and adhere to the "core values of socialism." Providers are also required to tag content generated using GenAI in accordance with the provisions provided by the regulations on the Administration of Deep Synthesis of Internet Information Services, which came into effect in January 2023. For training models, the service providers are required to obtain personal data with consent or under situations prescribed by the law or administrative measures and take measures to increase the quality of training data. 

• Later in August, China gave approval for the first batch of GenAI services to be publicly rolled out, allowing companies including Baidu and SenseTime to offer GenAI services to the public, competing with global players like OpenAI and Microsoft. 

• In September, the US federal government issued new guidelines for its employees using AI tools at work to ensure responsible and unbiased AI use. Treasury Board President Anita Anand emphasized that the government will closely monitor AI usage to mitigate potential biases or discrimination. Departments are required to recognize content generated through GenAI, inform users about their interaction with an AI tool, record decisions, and have the capability to offer explanations when AI tools are employed in decision-making processes.

• The Canadian federal government also introduced a voluntary code of conduct for GenAI amid concerns about its rapid development and proliferation. Innovation Minister François-Philippe Champagne unveiled the code at the All In artificial intelligence conference in Montreal, where several Canadian tech companies including Cohere committed to adopting it. 

• The UK's Competition and Markets Authority in September established seven specific principles centered on Foundation Models such as GPT-4 and Llama-2. The principles include ensuring accountability for AI outputs, guaranteeing broad access to essential resources like chips and training data, promoting diverse business models, providing businesses with model usage choices, ensuring flexibility and interoperability between models, preventing anti-competitive behaviors, and emphasizing transparency regarding the risks and limitations of GenAI content.

• During a lawsuit against the US Copyright Office, US District Court Judge Beryl A. Howell, in August, ruled that artwork generated by AI cannot be copyrighted, emphasizing that copyright has traditionally been granted to creations with discernible human influence. The future of US copyright law and AI remains uncertain, given the growing number of legal cases in this area against companies like OpenAI and Meta for alleged data scraping practices as well as software piracy claims against Microsoft, GitHub, and OpenAI related to data scraping.

Stages of regulations in major countries/jurisdictions as of Q3 2023

Source: Compiled by SPEEDA Edge based on Reuters and other sources

Other industry updates: Voluntary commitments from major players for safe and responsible development 

Analyst take: The voluntary establishment of an industry body for Frontier AI was likely prompted by the recent voluntary commitments on safe AI development made to the White House, highlighting its commitment toward this goal while reducing the probability of future litigation. 

• In July, OpenAI, Microsoft, Google, and Anthropic formed the Frontier Model Forum, an industry body to ensure the safe and responsible development of advanced AI models. The forum's mission includes ensuring AI research is safe, identifying best practices for responsible development and deployment of frontier models, collaborating with various stakeholders to share knowledge about trust and safety risks, and supporting efforts to develop applications that can help address major societal challenges.

• The Biden administration also gathered "voluntary commitments" from OpenAI, Anthropic, Google, Inflection, Microsoft, Meta, and Amazon to pursue common safety and transparency objectives. The commitments include internal and external security tests of AI systems before release, sharing information on AI risks and mitigation techniques, investing in cybersecurity, facilitating third-party discovery and reporting of vulnerabilities, developing robust watermarking for AI-generated content, reporting AI systems' capabilities and limitations, prioritizing research on societal risks, and developing AI to address significant societal challenges.

Funding: Foundation Models leads in dollar terms, with GenAI players raising the most rounds

• During Q3 2023, the Foundation Models segment attracted the most funding in dollar terms at ~USD 2 billion, despite the GenAI Applications segment leading in the number of funding rounds (25). 

GenAI Applications

Analyst take: This quarter reported a decline in funding compared with the previous three quarters. While it is too early to tell, this may be a sign of diminishing enthusiasm within the industry due to prominent tech giants' substantial influence and the realization that GenAI Applications might not yet be prepared for widespread adoption.

• During Q3 2023, 24 disruptors raised ~USD 270 million across 25 rounds in disclosed funding, recording a dip of ~56% compared with the previous quarter. The education, research, and finance segment attracted the bulk of funding during the quarter (~28% of total funding). The highest funding amount recorded by a disruptor during the quarter was AI character developer Inworld AI raising USD 50 million in August.

• Companies in the business process improvement segment accounted for ~21% of the funds raised during the quarter, followed by the entertainment segment at ~19%.

• In terms of the number of funding rounds (25), the education, research, and finance segment and business process improvement segment contributed equally with six funding rounds each, with an average deal value of ~USD 13 million and ~USD 9 million, respectively. However, the highest average deal value was recorded by the entertainment segment with USD 25 million.

• Seed funding was the most common form of funding during Q3 2023 (~64% of all funding rounds). However, early-stage funding dominated in dollar terms, contributing 41% of total funds raised. Notably, no growth-stage funding rounds were recorded.

• 16 startups raised seed funding during the quarter, with Canadian text-to-image solutions developer Ideogram drawing the highest infusion (USD 16.5 million), followed by employee experience platform Atomicwork (USD 11 million). Six startups raised early-stage funding, with Darrow (USD 35 million) and Harmonya (USD 20 million) leading.

Foundation Models

Analyst take: The Foundation Models space appears to be becoming progressively saturated, with major tech companies vying for dominance. A case in point is Amazon’s investment in Anthropic, which would allow Amazon to strengthen its position in the rapidly growing AI industry by adding more resources and capabilities. This also reflects Amazon's commitment to AI development, positioning itself against rivals like Microsoft and Google. 

• During Q3 2023, five players raised ~USD 2 billion across six rounds in disclosed funding (~22% less than the previous quarter, yet ~58% more than the total funds raised throughout 2022). The funding was split between players in the LLM segment (accounting for the bulk of ~89%) and fine-tuned language models (the remaining ~11%).

• The highlight during the quarter was Amazon’s investment of USD 1.25 billion in Anthropic, for a minority stake of the company, with an option to invest up to USD 4 billion. This round aligns with its previously announced intentions from April of raising a total of USD 5 billion over the upcoming two years.

• In addition, both AI21 Labs and Imbue achieved unicorn status with their USD 155 million Series C round (valued at USD 1.4 billion) in August and USD 200 million Series B round in September, respectively. Interest in healthcare-related LLM developers continued, with Hippocratic AI raising another USD 15 million in July (following a USD 50 million round in May 2023).

• On a separate note, OpenAI is looking at selling shares, potentially increasing its valuation from USD 29 billion to an estimated USD 80 billion–90 billion.

• Growth funding and unknown rounds were among the most common types of funding during the quarter (two rounds each). 

GenAI Infrastructure

Analyst take: GenAI Infrastructure, a critical component of the AI landscape, saw a remarkable uptick in Q3 2023. This underscores the continued vitality and expansion of the GenAI Infrastructure industry.

• Funding in the GenAI Infrastructure space picked up in Q3 2023, with 16 disruptors raising ~USD 913 million across 16 rounds in disclosed funding. This was 5x more than the funding reported in Q2 2023 and also the highest aggregate quarterly funding raised since Q4 2021. The data storage and retrieval segment contributed ~58% of funding during the quarter, with Databricks’ infusion of USD 504 million in a series I round marking the highest funding amount recorded by a disruptor during the quarter in the space. 

• Companies in the hardware infrastructure segment accounted for ~17% of the funds raised during the quarter, followed by the integrated LLMOps solutions segment and the model development and training segment accounting for ~11% each.

• In terms of the number of funding rounds (16), the integrated LLMOps Solutions segment accounted for ~31% of the total rounds, with an average deal value of ~USD 20 million, and model development and training accounted for ~19% of the funding rounds, with an average deal value of ~USD 33 million.

• Seed was the most common form of funding during Q3 2023 (seven rounds); however, the growth category dominated in dollar terms, contributing 64% of total funds raised (the Databricks round contributed ~86% of the category total).

• Other notable rounds during the quarter were a USD 110 million series B round by hardware infrastructure provider d-Matrix, a USD 85 million series C round by codeless AI infrastructure platform Pixis, and a USD 50 million (venture round) raised by LLMOps solutions provider Weights & Biases.

Repeat funding rounds during 9M 2023

Analyst take: Repeat rounds in a short timeframe indicate sustained investor interest and a push to strengthen expertise and product deployment. For instance, Anthropic reported four funding rounds within the first nine months of 2023. 

• 88 companies had one funding round during 9M 2023, with the majority being in the GenAI applications space (55 companies), followed by GenAI Infrastructure (24) and Foundation Models (nine). 

• Anthropic was the only company to report four funding rounds during the period (~USD 2.1 billion), while MindsDB (~USD 47 million) and Tavus (~USD 30 million) reported three funding rounds each.

Notable funding rounds and investors

Product updates: Driving advancements in product features, new models, and improving efficiency of development efforts

GenAI Applications

Analyst take: When GenAI chatbots like OpenAI's ChatGPT were introduced, their capabilities were quite basic, mostly limited to engaging in natural language conversations. However, both disruptors and incumbents introduced advanced features to chatbots during the quarter, including voice, image, and even coding capabilities. Additionally, industry players are expanding their reach globally and enhancing their chatbots with web browsing capabilities, adding versatility to their products. As chatbots become more advanced, they hold the potential to transform various industries such as customer service, content creation, and even software development. This trajectory is likely to redefine the GenAI landscape.

• We tracked 26 product updates during Q3 2023 across 12 disruptors. 

• Most of the updates were in the conversational content segment. The majority of the updates were by OpenAI (nine), which encompassed some major updates to its chatbot, including the launch for enterprises and rolling out image and voice capabilities. It also reintroduced web browsing capabilities in September after temporarily disabling them in July. Other chatbot providers like Perplexity and Anthropic were seen keeping up with new product launches offering advanced features, particularly coding capabilities. 

• In other segments, two companies launched consecutive updates during the quarter. One was Forethought (in the business process improvement segment), which launched an enhanced version of its Solve Email product, launched an open-source developer platform “AutoChain,” and launched “Autoflows” to build workflows with natural language. Stability AI (in the design, publishing, and digital assets segment) launched a sketch-to-image tool “Stable Doodle” and “Stable Audio” to generate music and sound. 

• Other companies that launched new products during the quarter are Omneky (launching the Creative Generation Pro tool), Got It AI (AI assistant Agent Copilot for customer service and sales), Persado (lEssential Motivation to help create emotionally engaging messages), Writesonic (Audiosonic to turn text into speech for people with disabilities), and HeyGen (a tool to translate videos).

• During Q3 2023, key incumbents also launched several GenAI-enabled products. In July, Google announced the expansion of its chatbot Bard, making it available in 40+ languages along with several new features like customized tone and style, vocalized responses, and enhanced productivity features. The same month, Microsoft upgraded its Bing Chat with visual search functionality, expanding to Chrome and Safari for select users, and launching Bing Chat Enterprise. The following month, it announced Bing Chat availability on third-party browsers, including mobile devices.

• Google followed the lead of disruptors launching a browser-based development environment “Project IDX” in August, expanding its coding capabilities followed by Amazon launching the enterprise tier of its coding product CodeWhisperer, allowing companies to integrate their internal codebases. Meta announced a major update in September, marking its entry into the AI chatbot space with its own assistant and a set of AI characters integrated into WhatsApp, Instagram, and Messenger. 

Foundation Models

Analyst take: Players in the space continue to launch new models with varying capabilities. A significant development is the expansion of GenAI beyond the English language, as Chinese and Korean players enter the market with native-language Foundation Models.

• OpenAI was again in the forefront, announcing the general availability of its GPT-4 model through an API. It also introduced a fine-tuning option to the GPT-3.5 Turbo model and released DALL-E 3, upgrading its text-to-image tool to offer improved context understanding.

• Multimodal model developer Stability AI had a particularly active quarter, introducing several new models across various domains including two LLMs: FreeWilly1 and FreeWilly2, capable of answering complex questions in specialized domains, Diffusion XL 1.0, an advanced text-to-image model, StableCode, an open LLM to assist users in generating programming code, and StableLM Alpha, a language model designed for Japanese speakers. Another multimodal model developer Midjourney released new features to its existing models offering features to expand images through prompts and enhance upscaled images. 

• Other disruptors active during the quarter were MosaicML launching new model MPT-7B-8K, an open-source LLM with 8k context length, AI21 Labs launching a tool (Contextual Answers) to increase efficiency and accuracy of information queries, and Deci, which launched three new models: DeciCoder, an open-sourced LLM for generating programming code, DeciDiffusion 1.0 a text-to-image model, and DeciLM 6B a generative text model, alongside its inference Software Development Kit called Infery LLM. 

• Incumbent activity slightly surpassed that of disruptors in Q3 2023, with 15 updates tracked across 9 incumbents. Meta was the most active during Q3 2023, with several launches, including multimodal model CM3leon, GenAI audio and music tool AudioCraft, SeamlessM4T that can translate and transcribe almost 100 languages across text and speech, and Code LLaMA, an LLM that uses text prompts to generate and discuss code.

• Notably, incumbents outside the US reported significant activity developing native-language Foundation Models. For example, Chinese tech giant Alibaba released several Foundation Models including LLMs Qwen-7B and Qwen-7B-Chat and multimodal models Qwen-VL and Qwen-VL-Chat, which can interpret both English and Chinese text and images. Other Chinese players Tencent (Hunyuan) and Ant Group (Zhixiaozhu 1.0 and Zhixiaobao 2.0) also launched Foundation Models. South Korean player SK Telecom Introduced "A" (also known as "A Dot''), the world's first Korean LLM, as part of its plan to become a global AI company. 

• Other incumbents active during Q3 2023 were Salesforce releasing XGen-7B, an LLM that supports longer context windows, Alphabet (Google) launching RT-2, a vision-language-action model that uses web and robotics data to guide robotic control, and Adobe launching the commercial version of Firefly models.

GenAI Infrastructure

Analyst take: An inherent limitation of GenAI technology has been its substantial operational costs, encompassing model training expenses and resource-intensive processing. As the GenAI landscape progresses, businesses are actively exploring avenues to enhance efficiency through hardware innovations that facilitate on-chip (edge) processing. This approach could offer a range of advantages, including reduced power consumption, accelerated processing, and secure data handling. 

• We tracked 24 product updates during Q3 2023 across 21 disruptors.

• A major interest for disruptors was introducing hardware solutions to enable on-chip (edge) processing achieving benefits like reduced power consumption, faster processing, and improved data security. In this regard, several disruptors including VSORA, Kneron, Esperanto Technologies, and MongoDB launched new products.

• Several disruptors including Datastax, Redis, and MongoDB launched new products to improve vector search capabilities to offer enhanced model capabilities, which enables developers to aggregate and filter data more efficiently, enhancing semantic information retrieval and reducing errors in AI-powered applications.

• Model monitoring was another area of interest, where Arthur, Appen, Arize AI, and TruEra launched solutions for LLM evaluation and troubleshooting.

• Incumbents were as active, with 19 updates tracked across 10 players. NVIDIA was the most active, with five product updates: (1) the announcement of its cloud-based AI supercomputing service, DGX Cloud, (2) the announcement of an updated version of its GH200 Grace Hopper superchip (to be available in Q2 2024), (3) the launch of RTX workstation graphics processing units (GPUs) optimized for GenAI content creation, (4) the launch of  AI Workbench, a tool that allows developers to build GenAI models on PCs, and (5) the announcement of plans to release TensorRT-LLM, an open-source software to enhance the speed and efficiency of LLM inference.

• Microsoft announced a public preview of Vector Search in Azure Cognitive Search and launched open-source library based TypeChat in July. The company also announced geographically expanding Azure AI infrastructure in August. Google launched AlloyDB AI, a set of integrated capabilities within AlloyDB for PostgreSQL, supporting developers in creating GenAI Applications using their own data, and updates to its Vertex AI platform, including improved AI models for text, image, and code generation, and added third-party models from players like Anthropic and Meta.

• IBM introduced an analog AI chip designed for efficient deep neural network computations and WatsonX Code Assistant Z that converts COBOL to JAVA, set to be available in Q4 2023. On the other hand, Intel announced that its forthcoming "Meteor Lake'' laptop chip, set to release in December, will have the capability to run GenAI chatbots directly on a laptop without relying on cloud data centers.

Partnerships: Product collaborations to improve accessibility of models and infrastructure

GenAI applications

Analyst take: Established companies in the space like Jasper and Typeface were seen strategically collaborating with incumbents to expand the accessibility of their solutions. Notably, the majority of incumbent partnerships recorded during the quarter were customer partnerships, potentially signaling increasing adoption of these solutions.

• We counted 12 partnerships in the GenAI Applications space during Q3 2023, across seven disruptors and three incumbents. 

• Among the disruptors six of these partnerships were product collaborations, which included 1) OpenAI’s partnership with American Journalism Project (AJP) to support local news through AI, 2) marketing platform Jasper’s partnership with Google Workspace, Webflow, Zapier, and Make to offer access to Jasper tools in the partnered platforms, 3) marketing platform Typeface’s partnership with Alphabet (Google Cloud) and GrowthLoop to launch a unified marketing solution, 4) voice generation platform ElevenLabs’ partnership with D-ID to integrate ElevenLabs' AI voice technology into D-ID's Creative RealityTM studio, 5) business process platform Glean’s partnership with DoiT to leverage DoiT's expertise in cloud operations to manage cloud costs, and 6) Anthropic’s partnership with BCG to provide BCG's clients with access to Claude 2 and other AI technologies.

• Legal solutions provider Harvey recorded a customer partnership with Macfarlanes, which integrated GenAI into its client services, as an early adopter in the legal sector's growing use of GenAI.

• Microsoft accounted for three of the five incumbent partnerships with a product collaboration with Box to launch a plugin for Microsoft 365 Copilot, allowing customers to synthesize and summarize shared Box documents in Teams and two customer partnerships with the Government of Japan and Lumen Technologies.

• Baidu and Fujitsu also reported customer partnership with Great Wall Motors incorporating Baidu's Ernie Bot, enabling enhanced interaction between the driver and the vehicle, and Fujitsu piloting its retail marketing content solution at Aruk Mitajiri Store, a supermarket run by Marukyu Co.

Foundation Models

Analyst take: With their presence firmly established, disruptors in the field are now directing their attention toward enhancing their models. They are actively forming partnerships to access additional training data and to refine the fine-tuning capabilities of these models. In parallel, companies operating in this sector, including both disruptors and incumbents, are actively seeking partnerships to broaden the accessibility of their models.

• In the Foundation Models space, disruptors and incumbents were equally active, with eight and nine partnerships recorded, respectively. 

• Disruptors including OpenAI, Stability AI, and AI21 Labs were seen forging partnerships to further improve their models. Notably, in July, OpenAI partnered with both Shutterstock and Associated Press to train its models on image data and news stories, respectively. It also partnered with Scale AI in August to offer fine-tuning tools for OpenAI's model GPT-3.5, allowing users to refine GPT-3.5 to create tailored models for their unique business requirements. Stability AI on the other hand focused more on improving the efficiency and speed of models by partnering with NVIDIA and WEKA. 

• Disruptors were also focused on expanding the availability of models; for example, AI21 Labs partnered with Google to make AI21 Studio LLMs available on Google Cloud Marketplace.

• Meta dominated the incumbent space in terms of partnerships, mainly to expand the availability of its models. The company partnered with Microsoft, Alibaba, IBM, and Amazon to offer access to its Llama 2 family of models on these platforms. On a separate note, Meta also partnered with Qualcomm Technologies to bring on-device AI processing of Meta's models by running them directly on devices and MediaTek to to accelerate AI application development across various edge devices. 

GenAI Infrastructure

Analyst take: As established players in the cloud infrastructure domain continue to broaden their expertise, companies are finding value in collaborating with them to develop GenAI Applications. This collaborative effort serves dual purposes, enabling companies to not only offer GenAI solutions to their clients but also to use these technologies for internal process enhancements and operational efficiency. Such partnerships could be a driving force behind the expansion of the entire GenAI ecosystem.

• We tracked 11 disruptor partnerships in the GenAI Infrastructure space. Several disruptors including Pinecone, Snorkel AI, Databricks, Redis, and DataRobot, expanded the availability of their offerings in cloud platforms like Microsoft Azure, Google Cloud, and Amazon Bedrock by partnering with them. Pinecone also partnered with Datadog to offer observability and monitoring capabilities for vector databases and Generative AI Applications. 

• Incumbents recorded 33 partnerships during Q3 2023. Most of these involved various companies teaming up with cloud infrastructure providers like Microsoft Azure, Google Cloud, and Amazon Bedrock to use the cloud-based infrastructure and Foundation Models available to create GenAI solutions either for customer products or to improve internal operations. One example of the former is Cognizant’s partnership with Google to develop healthcare solutions on Google Cloud's GenAI technology. An example of the latter is Genpact partnering with Microsoft to offer Genpact's global teams access to Azure OpenAI Service to enhance employee productivity and operational efficiency. 

• NVIDIA entered several partnerships to offer hardware solutions for improved performance. For example, it partnered with Dell to integrate Tensor Core GPU with Dell's AI software and data storage and with Hugging Face to introduce “Training Cluster as a Service,” aimed at streamlining the development of GenAI models for businesses by integrating NVIDIA's DGX Cloud supercomputing resources. Another notable update was NVIDIA’s partnership with two Indian tech giants, Tata Group and Reliance Jio, to develop LLMs trained on India's diverse languages and create AI infrastructure using NVIDIA’s supercomputer solutions. 

Incumbent engagement in the startup ecosystem

Analyst take: Incumbents are entering strategic partnerships with a twofold objective; first, to open their infrastructure for collaborative application development, and second, to facilitate the integration and visibility of their partners' products within their platforms. These partnerships not only enable partners to extend the reach of their products but also help incumbents to diversify and improve their platform offerings, fostering a dynamic ecosystem of mutually beneficial solutions.

• During Q3 2023, several incumbents invested in startups, the most prominent being Amazon’s investment in Anthropic. Both Google and NVIDIA invested in AI21 Labs. In the infrastructure space, NVIDIA invested in MindsDB as well as Databricks, while Microsoft invested in d-Matrix. IBM was also active as an investor with interest in Hugging Face. 

• In terms of partnerships, Microsoft, Google, and Amazon were actively entering product and customer partnerships to offer cloud services to application developers. On the other hand, most of NVIDIA’s partnerships were focused on offering necessary hardware infrastructure. Incumbents also partnered with model developers and infrastructure services providers to make the disruptors’ products available on incumbent platforms.

Note: The incumbent activities include both partnerships and investments. Refer the relevant appendices for more details

Source: Compiled by SPEEDA Edge from multiple sources

M&A: OpenAI reports first acquisition in its seven-year history

• In August, OpenAI announced the acquisition of Global Illumination, a developer of AI-based creative tools, infrastructure, and digital experiences, its first public acquisition in its history of seven years. Global Illumination has demonstrated a record of innovating tools and transforming digital experiences, making its expertise a fit with OpenAI's mission to continually advance the frontiers of GenAI.

• During the same month, IBM completed the acquisition of Apptio, integrating its FinOps solutions, including ApptioOne, Cloudability, and Targetprocess, with IBM's automation tools like Turbonomic, AIOps, and Instana. IBM also planned to enhance its Watsonx AI and data platform by incorporating Apptio's vast pool of anonymized IT spending data to support the development of AI and Foundation Models.

Appendices

Appendix 1

Appendix 2 

Appendix 3