BUSINESS

CLUB

INDIAN
CASE
CHALLENGE
2024

PRE - CASE
 BUSINESS
CLUB

Unravelling of OpenAI
The saga kicked off on the 17th of November when the Board of Directors
abruptly terminated CEO and co-founder Sam Altman during a video call that
Altman assumed would be an uneventful check-in. This bombshell and tragic
news became public rapidly, with Microsoft discovering it around the same
time as the entire world.

OpenAI has emerged as Silicon Valley's hottest and most valued AI startup of
recent years, thanks to the viral user growth of its ChatGPT chatbot. Propelled
by ChatGPT's runaway popularity, OpenAI now boasts over 100 million weekly
active users interacting with its flagship conversational agent.

6 Nov 17 Nov 18 Nov

OpenAI makes major Sam Altman is sacked by the COO Lightcap, attributes it to
announcements on board, Greg Brockman resigns in communication issues, Mira Murati
Devday solidarity appointed as the CEO of OpenAI

Altman returns as
the CEO of
OpenAI

21 Nov

21 Nov 20 Nov 19 Nov

Ilya Sutskever expresses Ex-Twitch CEO Emmett Mircrosoft announces hiring
regret over Sam’s Shear announced new Sam & Greg to lead a new
dismissal CEO division

Exhibit 1 - Timeline of events

1
 BUSINESS
CLUB

Altman's dismissal came as a massive shock across the technology industry

and internally at OpenAI. In an act of solidarity, President Greg Brockman
resigned on the same day that Altman was let go. Initial speculative reports
suggested that a strained relationship and communication breakdown
between Altman and Chief Operating Officer Brad Lightcap may have been
an underlying factor behind Altman's removal.

Things seemed stable for a short time, but on November 20th, unexpected
events unfolded, shaking up the situation. OpenAI revealed they had selected
former Twitch CEO Emmett Shear as the company's next leader. Over 500
staffers signed a petition demanding Altman's return, or they would resign en
masse. Even Chief Scientist Ilya Sutskever voiced doubts over the board's
decision-making regarding Altman.

2,000

1,500

1,000

500

0
11/2022 12/2022 01/2023 02/2023 03/2023 04/2023 05/2023

Exhibit(2) - Monthly visits to OpenAI

2
 BUSINESS
CLUB

By November 21st, after internal discontent, OpenAI declared they had agreed
that Altman would be reinstated as Chief Executive Officer under a partially
new and reconfigured board. This dramatic reversal was completed over a
week of unprecedented turmoil and volatility surrounding the company. In
the end, the ordeal reaffirmed Altman as the leader.

Altman's removal coincided closely with internal debates concerning the safe
use of AI within the company. A report by The Information indicated internal
disagreements and conflict among employees regarding the safety measures
being undertaken in the development of AI. Despite the turmoil, OpenAI
launched an assemblage of new features on 6th November and unveiled their
latest innovations, garnering global coverage in an event called DevDay.

DevDay

“92% of Fortune 500 companies are using our solutions, some two million
developers are building on top of the technologies via our API, and ChatGPT
has 100 million weekly users, growing only by word of mouth”, said the CEO
Sam Altman, in high spirits. OpenAI’s DevDay 2023 marked a significant
milestone in AI; it was explicitly crafted for developers. The introduction of the
Assistants API represented a pivotal tool for developers crafting sophisticated
and interactive AI-driven applications.

3,000

2,500

2,000

1,500

1,000

500

0
2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032

Exhibit(3) - Market size of AI industry (USD Billions)

3
 BUSINESS
CLUB

The newly announced tool would significantly reduce development time and
costs, allowing a much smoother API integration. The firm announced a more
economical and efficient GPT4 Turbo and, most significantly, GPTs. With the
rollout of GPTs, users could now train their custom-trained chat assistants like
a GPT to plan and update their calendars in real-time. They would be
available as plug-ins in the ChatGPT window and can be published for
everyone to use; many of them could be built with little or no code.

GPTs open up opportunities for innovators to build them with ease and share
them with the entire user base. To facilitate this, OpenAI introduced GPT store
and GPT builder, making the LLM more accessible.

40
30
20
10
0
Hardware Software Services

Exhibit(4) - Artificial Intelligence Market Size By Solution, 2022 (%)

GPT Store
The GPT Store is an innovative, collaborative AI marketplace featuring custom
GPT models. It aims to democratise access to AI models, enabling developers
to utilise pre-trained models and contribute and share their models within a
collaborative ecosystem. The GPT Store and GPT Builder work hand in hand
to make customised AI models accessible. These AI models can be
customised to fit user needs. For instance, an educator could use the GPT
builder to create a model to teach complex scientific concepts. This process
of building models has been made significantly straightforward, thanks to the
GPT builder.

4
 BUSINESS
CLUB

The GPT builder takes instructions in a conversational tone, enabling the user
to train the model effectively. The builder thus simplifies the process of
prompt engineering the model by taking the user's simple instructions,
additional knowledge and functionalities. Once the user is satisfied with their
model, they can publish it on the GPT store for the world to use.

Industry leaders have speculated that the GPT Store will only be accessible to
paid users initially, though that may change. Subscribers will build the
models through prompting and be able to push them onto the platform.

Exhibit (5) - OpenAI is now world’s 10th largest Unicorn

Revenue Sharing
The GPT Store's revenue-sharing aspect is intriguing and critical. While the
initial focus is on user engagement, there’s room for alternative approaches.
Each approach will present its unique set of circumstances to Open AI.

5
 BUSINESS
CLUB

While keeping it accessible to premium users can incentivise people to buy

their subscription, there’s always a tradeoff as OpenAI loses its share of new
subscribers who would've tuned in to try the all-new and customised models.
Furthermore, this new revenue-sharing model would require a compromise
between developer incentives and company profits.

Determining the revenue-sharing model that aligns with OpenAI's

commitment to advancing AI development is crucial. Should access to the
powerful GPT Builder be restricted to paid users, or should some level of
functionality be available to everyone? OpenAI has aimed to incentivise
developers by promising monetisation. However, the choice to restrict access
exclusively to premium users warrants justification. While opening models to
the general public might increase traffic to OpenAI, the incentive for
premium subscriptions may outweigh the advantages of broader
accessibility.

Potential Pitfalls
The GPT Store could also face several other problems, oversaturation being a
major concern. Due to ease of creation and the limited revenue per user, the
store could quickly be flooded with custom GPTs. This could cause several
issues:

1. An influx of low-quality GPTs significantly hinders user experience and

technology acceptance.
2. Increased competition among similar GPTs, causing financial strain for
developers and taking away existing apps and SaaS solutions' user base
and revenue.
3. Users might need help discovering relevant models, leading to fatigue and
frustration navigating the vast array of available GPTs.

6
 BUSINESS
CLUB

AI-Specific Chips
Another possible concern with the GPT Store is the high influx of users, which
could strain the computational resources of OpenAI, which are stretched thin
to begin with. Computational resources, specifically AI chips, have become
important in AI development. These chips offer notable performance and
efficiency gains over conventional CPUs and GPUs because they are
specifically engineered to meet the demands of AI workloads.

Companies like Microsoft and OpenAI are creating specialised AI chips and
are driving this revolution. Microsoft has released the Azure Maia 100 and
Cobalt 100 chips, designed primarily for general-purpose computing and AI
training. In the meantime, OpenAI, led by Sam Altman, is researching and
developing the TIGRIS chip, which will challenge Nvidia's hegemony in the AI
hardware market.

200
Performance/Watt Relative to CPU or GPU

150

100

50

0
Total Perf./Watt GM Total Perf./Watt WM Incremental Perf./Watt GM Incremental Perf./Watt WM

Exhibit (6) - Performance comparission of CPU/GPU

7
 BUSINESS
CLUB

250

200

150

100

50

0
2032 2031 2030 2029 2028 2027 2026 2025 2024 2023 2022

Exhibit (7) - Artificial Intelligence (AI) Chip Market Size (USD Billion)

A number of startups have also sprung up in recent times to capitalise on this

relatively new industry. Few of them are given below.

Cerebras: Founded in 2016 by Andrew Ng, a renowned AI expert, Cerebras is

pushing the boundaries of AI hardware with its wafer-scale chip architecture.
Unlike traditional chips with billions of transistors on a single silicon wafer,
Cerebras uses hundreds of interconnected wafers to create a massive chip.
This "one-chip computer" boasts 1.4 trillion transistors, overpowering even the
most powerful GPUs. Cerebras chips can handle massive datasets. They
provide a higher level of efficiency, opening new avenues to explore natural
language processing, protein folding, and climate modelling.

Key Features:
Wafer-scale architecture with 1.4 trillion transistors
Massively parallel processing for large datasets and complex models
Reduced power consumption compared to traditional AI hardware
Focus on high-performance computing and scientific research

Problems:
Chips are not readily available due to complex architecture and
manufacturing for wafer-scale processors.
Excels at raw processing power are not optimised for specific AI tasks like
deep learning.

8
 BUSINESS
CLUB

Graphcore: While Cerebras focuses on sheer scale, Graphcore takes a different

approach with its "intelligent processing unit" (IPU) architecture. These
particular chips are made to perform the calculations required by graph
algorithms and deep learning. IPUs offer substantial performance gains due
to their specialised cores and memory, which optimise data flow and
communication, setting them apart from CPUs or GPUs.Pfizer, BMW, and
Facebook already use Graphcore's IPUs for various AI applications, from
financial modelling to drug discovery.

Key Features:
IPU architecture optimised for deep learning and graph algorithms
High-bandwidth memory and specialised cores for efficient data flow
Scalable architecture for large-scale AI deployments
Focus on practical AI applications in various industries

FPGA
9.6%

ASIC
15.1%
GPU
42.4%

CPU
32.8%

Exhibit (8) -Global AI chipset revenue by market share

9
 BUSINESS
CLUB

Problems:
Excels at specific tasks but limited scalability for massively parallel tasks.
It uses its domain-specific language (Poplar), which might bug developers.

This rapidly rising segment of the AI specific chip industry has opened up a
new facet. The success of many of these large AI firms will depend on how
they can complement their state of the art software with specialised
hardware which will boost the pace of development of AI.

Conclusion
OpenAI is undoubtedly at the forefront of AI development owing to the
unprecedented adoption of ChatGPT and DALL-E, while also being in a
continuous tussle of capital, regulation and innovation. Many firms, although
not as prominent as OpenAI, are capitalising on the booming AI market
fueled by heavy demands from individuals and enterprises alike. GPTs and
GPT stores would mark a critical step in determining OpenAI’s market
position for building the most pragmatic solutions over a hyper-competitive
landscape. In order to sustain the research the firm is best known for,
developing AI-specific chips could be a viable option, but it comes not
without its own drawbacks and challenges. Whether OpenAI navigates
through these barriers successfully or not is unclear. However, what the future
beholds for AI is indeed enticing.

10
 BUSINESS
CLUB

Questions
1) GPT Store with the GPT Builder has the potential to revolutionise chatbots,
offering a level of customisation and freedom to its users never seen before
in the industry. However, it also comes with a number of questions for
OpenAI. The company can employ several different methods to monetise
the GPT Store based on its long-term goals. Keeping various industry factors
in mind, construct an appropriate revenue model for the GPT Store,
exploring the proposed model's financial feasibility and a go-to-market
strategy to stand out among its competitors.

Also, recommend strategies for OpenAI to address the potential for

oversaturation in the GPT marketplace and ensure that the platform fosters
the development of high-quality, specialised, and sustainable GPTs that
meet the diverse needs of users.

2) OpenAI’s recent venture into the semiconductor market opens up several

opportunities for the company along with several challenges. Braving it out
alone in an industry requiring huge capital investments could be a major
challenge for the company. Keeping this in mind, suggest a strategy for
OpenAI to onboard a firm or startup in the AI chip design industry and the
feasibility of such an onboarding. Justify your suggestion considering various
metrics such as economic and market synergies, goals, etc, and state the
advantages of such a move for both sides.

11
 BUSINESS
CLUB

Rating Criteria and Qualification Requirements

For the final round, the teams will be evaluated on the basis of the following
criteria:
Quality of Analysis
Creativity
Feasibility of the Solution
The overall Presentation of the Case

Note
Teams must submit their presentation by 11:59 PM, 3rd Jan, 2024.
Submission should be in PPT format only(maximum 15 slides excluding the
introduction, final slide, and Appendix) with an accompanying PDF copy of
the same presentation.
The solutions should be mailed to the [email protected] with
attachments as .pptx and .pdf.
Relevant information can be added for the analysis from the web or other
suitable resources.
Any form of plagiarism will be heavily penalized.
The inclusion of an executive summary slide is mandatory.

Copyright Notice
All rights reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, electronic,
mechanical, photocopying, recording, or otherwise, without the prior written
permission of Business Club, IIT Kharagpur.

This case is the intellectual property of Business Club, IIT Kharagpur.

For further information, visit our website at https://icc.bclub.co.in/ or contact us

at our FB page https://www.facebook.com/bclubkgp.

For any further queries, feel free to call or Whatsapp:

1. Rajshekhar Upadhyay - 9137454490
2. Arnav Bansal - 9522280723

12
 BUSINESS
CLUB

References

Artificial Intelligence Market

Microsoft-OpenAI partnership
ChatGPT Operating Costs
Cloud Infrastructure cost comparison
The State of AI - Industry Report
Developments in AI
OpenAI: about the company
OpenAI dependance on microsoft
OpenAI products
OpenAI Company Profile
Sam Altman Dismissal Timeline
Gemini vs Chatgpt
Graphcore
Devday Announcement blogs
OpenAI current market position
OpenAI board/management
Global AI chipset Market
OpenAI - Generative AI market opportunities
AI Chip ventures
Cerebras’s chip development
Understanding the semiconductor structure
Graphcore IPU processors
AI roadmap
Google’s gemini stacking against OpenAI
Current state of the semiconductor industry
Sam Altman’s vision for future of AI
Classification of AI

13