Anthropic Launches Haiku 4.5: Smallest, Affordable AI Mode
As Anthropic launches Haiku 4.5, it’s the first time the US-based startup has updated a model in over a year. Could it all be to magnify AI’s appeal?
Before the question of real-world outcomes became the central cause of worry, the actual concern surrounding AI was its exponential pricing.
While there are sustained investment interests, the operational costs of running AI systems pose a blatant limitation. A limitation of its own development and scalability. And as a result, a roadblock to its own capabilities.
These economic bottlenecks are creating restlessness in the market. The costly demands that AI comes with will be a direct cause of slow development and may even limit accessibility. It could reposition the market’s focus towards monetizable AI infrastructures, not fundamental research.
This restriction influences the overall AI ecosystem. And additionally, investment.
To retain AI’s appeal, Anthropic has launched its most affordable and smallest AI model yet, Haiku.
Now that the AI race has gained some uniform momentum, tech businesses are searching for ways to combine innovation with affordability. They want AI systems that perform the same functions as any other advanced tool, but at a fraction of the cost.
The hardware used for AI development, from GPUs to computational resources, is the driving force behind the costs plaguing the AI industry. And this will only rise, i.e., as overall global energy consumption increases, the net demand will also skyrocket.
And over 30-40% of the overall number will be accounted for by AI.
This is precisely what Anthropic hopes to tackle with Haiku.
The US tech startup’s updated model, Haiku 4.5, is built using one-third of the costs of Sonnet-4. And one-fifteenth of Opus’. But it performs all tasks, including coding, as well as the other models.
In the early days, the selling point for businesses was talking up their most advanced and powerful AI models. But when clients would take a step back in caution of the roof-touching costs of using the best models, things had to change.
Companies have had to think small since then. And this is only a small step towards integrating different models- one that strategizes and one that does the grunt work.
To foster more efficient and productively innovative operations.
WPP & Google Partner to Transform Marketing Workflow – Ciente
As part of the agreement with Google, WPP will retain 300 million euros per year to invest in its AI future- remain a forerunner as it elevates client experiences.
CMOs and CEOs used to be misaligned about marketing’s role in a business’s growth and transformation.
The recent years have transformed this purview. Access to advanced technology has afforded tools to bridge the hidden cracks between marketing functions and the bottom line.
But today, with the whole picture in their hands, more business leaders are investing in it.
80% of growth leaders outperform their competitors. And they’re the ones who realize marketing’s true potential- its function as a growth accelerator.
This is the future that Google and WPP’s partnership plans to build upon.
They’ve expanded their five-year alliance to explore how marketers approach creative processes and transform the entire crux of marketing storytelling. The two giants aim to curate an integrated approach that empowers teams to enhance their quality of real-time personalization.
Google, through its AI capabilities, plans to help WPP revolutionize marketing as we know it. And take a giant leap beyond the traditional initiatives.
It’s a shared commitment to innovation. And a giant leap towards market-leading outcomes.
As a part of the agreement, WPP has made a $400 million spending commitment for Google tech. Some of the amount will be spent on integrating AI across its existing services, while some will be attributed towards AI investments made through WPP’s AI marketing platform, WPP Open.
“By delivering bespoke AI solutions and enabling hyper-relevant campaigns with unprecedented scale and speed, we’re accelerating innovation across every facet of marketing to drive unparalleled growth and impact,” chimed in Cindy Rose, WPP’s CEO.
This strategic alliance will establish new frontiers for WPP’s clients and what they can achieve with the revamped tech stack- the martech stack of tomorrow. The future of AI in marketing is now.
DGX Spark: Smallest AI Supercomputer to Democratize AI
Nvidia has harnessed complexity, there’s now an efficient supercomputer that sits right at your desk. And it is powered by AI.
Nvidia has launched the DGX Spark, the world’s smallest AI supercomputer. Jensen Huang personally delivered the first unit to Elon Musk at SpaceX’s Starbase in Texas.
He said, “Imagine delivering the smallest supercomputer next to the biggest rocket.” With a laugh. It is clear what Nvidia is signaling here, these are the next gen computation powers that will enable us to target the next frontier: space.
And Elon Musk is, once again, in the middle of it. Not for a rocket launch this time, but for an unveiling that might redefine what AI hardware looks like. A supercomputer that fits on a desk and connects seamlessly with existing systems.
Nvidia calls the DGX Spark a “rocket engine for AI.”
What DGX Spark really is
The Spark isn’t just a scaled-down version of the DGX systems. It’s designed to bring supercomputing-grade performance directly to researchers, startups, and developers who can’t access massive data centers.
It’s compact, modular, and powerful enough to train and deploy complex AI models locally.
Built on Nvidia’s latest architecture.
Tuned for generative AI workloads and multimodal models.
Designed for personal or lab-scale experimentation.
Why it matters
DGX Spark could shift how AI research is done. Instead of queuing for cloud access or HPC clusters, developers can now iterate faster and run high-intensity workloads from their own desks.
It’s not just a hardware release. It’s Nvidia reshaping access to AI compute, turning what was once infrastructure into an instrument.
The bigger signal
This is more than a delivery to Elon Musk. It’s a statement of intent from Nvidia. They’re not only building the backbone of global AI infrastructure. They’re making that power personal.
DGX Spark marks a transition point from cloud-scale AI to desktop-scale intelligence.
And if history is any guide, that shift changes everything.
LSEG-Microsoft Announces the Next Phase in Their Multi-Year Partnership
Microsoft’s cloud and LSEG’s analytical data make a surprising combination. Could this multi-year alliance transform the finance industry from the inside out?
Have you noticed how many places don’t require swiping your debit or credit cards today? Transactions went from ATM transactions to Venmo and RTPs. And even cards today have another option- to “Tap and Pay.”
Tech has brought an influx of transformations to the finance industry. It was digitization that changed the game.
The way people interact with banking services and transact has undergone serious vamping. Financial services, from B2B to D2C and from global to local levels, have become more secure, seamless, and instant.
But as the finance sector expands, the concerns have also risen.
There are more heaps of customer data. This is the cost of digitization. You can’t eradicate any of it. Amid the complexity, all data is a goldmine.
But how can fintech actually use it to their advantage? Cue in: AI.
This is precisely what Microsoft and LSEG’s multi-year partnership is built on.
According to the agreement, Microsoft is embedding its high-level AI capabilities across each financial workflow and layers of LSEG. It’ll inherently transform how LSEG’s customers access their data. And also establish a secure bridge for deploying agentic AI through an LSEG-managed MCP server.
“LSEG’s partnership with Microsoft is transforming access to data for financial professionals with cutting-edge, AI-driven innovation at scale,” states LSEG’s Group Chief Executive Officer, David Schwimmer.
The collaboration will facilitate a strategic deployment of agents across Microsoft Copilot Studio through LSEG data. Users can now access LSEG-licensed data and analytics directly through their workflows.
Additionally, users can develop sophisticated AI agents and integrate them within their workflows with proper governance control. And with compliant and secure customization at scale.
Given that the finance sector is highly (and complexly) regulated, innovations must adhere to all necessary regulatory compliance. And algorithmic bias and data security continue to be a concern. This is why most fintechs lack a robust regulatory framework in place.
Microsoft’s partnership with LSEG might be a much-needed step.
It connects LSEG-trusted content with Microsoft’s AI capabilities to remove any barriers to innovation across finance. As per the announcement, the strategy is built on LSEG’s strategy, LSEG Everywhere, creating a pathway to scale AI across financial services.
The unparalleled quality of LSEG’s AI Ready Content and taxonomies serves as the right foundational blocks for Microsoft’s Copilot to build upon.
Figma and Google Cloud Collaborate to Empower Creatives
Creative demands are changing as AI pushes the boundaries of creative expression. Can the Figma-Google Cloud partnership help fulfill these needs?
Stop waiting, start creating.
Figma and Google Cloud just reiterated their partnership.
The goal?
Instill Google’s fastest generative AI, the Gemini 2.5 Flash, directly into Figma’s design platform- a crucial deal for the platform’s 13 million monthly users.
The headline is all about speed.
Early tests showed a stunning 50% reduction in latency for Figma’s “Make Image” feature. What does that mean? When you ask the AI to generate or edit a visual, it happens twice as fast. For designers, that’s less time staring at a loading bar and more time in the flow.
This is about powering image generation “without breaking their flow,” said Figma’s CEO Dylan Field.
This AI integration eliminates those awkward pauses. You can now conjure up icons, mockups, or textured backgrounds almost instantly.
The speed originates from the Gemini 2.5 Flash model, which is optimized for quick responses. It’s what turns AI from a fun novelty into a genuinely essential tool.
But Figma isn’t stopping at speed.
The platform is also integrating other Google models: Gemini 2.0 and the super-realistic Imagen 4. It gives designers an entire toolkit. They can use simple text prompts to create high-quality images and even test out product functions before anything is built.
Google Cloud CEO Thomas Kurian noted the collaboration will help “push the design market forward.”
For the creative world, this move means less time wrestling with tedious workflows. It’s more time focused on big ideas, creative expression, and building the future.
And the next great design is now just a lightning-fast prompt away.
Total Addressable Market (TAM): Determine Your Business’s Viability
TAM is a dream- what if the entire world buys our solutions? But this thinking doesn’t sway investors. If chasing this high doesn’t lead to sustainable growth, what does?
“How big is this thing going to be?”
It’s the very first question every startup must face. And one that a lot of them skip out on.
They draw on ideas and business strategies, but the market size remains in question. Although it’s evident that they aren’t selling to the whole market, the markets they’re catering to are either too saturated, too microscopic, or don’t exist at all.
Existing data states that 70% of startups fail because they haven’t understood their market size. And this is why they’re missing out on a myriad of opportunities.
Your “amazing” product will fail to create an impact if you treat Total Addressable Market (TAM) as just a number in your pitch deck.
“A metric popularly leveraged by startups, entrepreneurs, and investors to evaluate the potential size of a market for a new product or service. It’s intended to represent the total potential demand for a product or service within a given market.”
Total Addressable Market is a mainstay in every startup’s pitch deck, presented as the potential revenue opportunity the company is pursuing. It’s the primary filter for investors, and demands that you present a hue, quantifiable market size that’s ripe for taking.
But investors aren’t swayed by this, given the current market conditions. What’s in large existing markets? It has become a blind spot- this is the main problem VCs now have with TAM.
Zeroing in Only on the Total Addressable Market Comes With Limitations.
Focusing on such massive markets is distracting businesses from actual opportunities- the creation of new markets.
Across this arena, TAM is negligible, barely attracting attention. It’s great for existing and well-defined markets, but for new or disruptive product categories, it’s a fluke.
The “top-down” approach to TAM analysis is taking the punch. It has come to be known as the unreliable pitch deck theater by the market.
What is the “Top-Down” TAM assessment approach?
The top-down approach starts at TAM- the total addressable market, i.e., the overall market size.
Then, they progressively narrow it down to the potential market size for the business or the eventual market share.
This comprises the target market segments for the business’s offerings out of the entire market (Serviceable Available Market, or SAM). And then you dive in even deeper, i.e., the percentage of market size that can be realistically obtained and serviced to (Serviceable Obtainable Market, or SOM).
TAM = the whole 10-inch pizza.
SAM = the slices you can eat.
SOM = the slices you’ll actually end up eating.
The calculations are often subjective and presumptuous. These numbers are based on industry reports, various macroeconomic factors, and market research data.
This makes the traditional TAM calculation a significant hurdle for business leaders.
These numbers could end up over-inflated. Focusing on a chunk of the market doesn’t automatically generate more opportunities. And the glimpse into a company’s vast potential (or the upper limit) that TAM offers is misleading.
Limitations of the traditional “top-down” TAM analyses
The traditional TAM framework is inherently unsuitable for creating new markets. In recent times, it has driven investor skepticism and led to startups undermining their true potential. And investors might easily end up passing over the company, missing out on the company’s potential within new markets that could be created.
VCs now want believable numbers that align with your product strategy, GTM framework, and final objectives. Only focusing on TAM introduces too many limitations-
1. “For truly innovative products, the market is not a fixed entity to be captured, but a dynamic one to be created.” In simple terms? Disruptive offerings create their own demand (see Uber, Apple, and Microsoft).
2. TAM is often a farce, a large number on paper to make a company look good. Do the presented numbers actually link to the business’s analytical and real-world capabilities?
Imagine a client you’re working with asserts that their TAM is around $60 billion. They’re convinced. But the number is quite ambitious and should be credible. So, their SOM turns out to be $600 million. This 100x overenthusiasm can put a hitch in the client’s business strategies and growth planning.
3. Top-down TAM analysis put together by a third-party source reeks of laziness and lacks credibility. Just because everyone around the world uses phones doesn’t mean all of them will purchase an iPhone. It’s become a norm for third-party providers to tie up existing categories to reflect a large number. But what do those pre-packed figures really demonstrate?
4. The market is extremely heterogeneous- with distinct buying behaviors, needs, preferences, and myriad variations of the same offerings. TAM or the overall market size is a subjective number that represents the potential demand for a company’s offerings. But with the market fragmented and chopped into segments, this number doesn’t consider the challenges of penetrating these segments.
For a cloud solution, there might be individual as well as enterprise buyers. This demands majorly distinct sales, marketing, and distribution approaches.
5. TAM is too static a number for the dynamic market of the 21st century. The market transformation cycles have elevated from a 10-year-long pace to just 2-3 years. It’s due to the increased flow of information, tech disruptions, the adaptive nature of business models, and dynamic customer preferences. And these factors are what should impact TAM, but its rigidity doesn’t have the room to consider these fluctuations.
“I hear these TAMs – we are attacking a $30B TAM. OK, that’s great. But where did that number come from? The single largest company that exists in what you define as your category is a $2 billion company. So, where is the other $28B? Where are these numbers coming from? And then you have to unpack it- is it software? Is it software services or something else? They are packing it in so it looks gargantuan. But you look at every precedent company and nothing is over $2B.”
The problem here is that the analytical capabilities of the traditional TAM framework are lacking.
What’s the solution here?
You lean on alternative methodologies to calculate TAM- ones that dive into the specifics to gauge the accurate market size.
TAM Analysis: Alternative Approaches to the Top-Down
A. The Bottom-Up Framework
TAM analysis with a bottom-up approach is more granular and specific. And it’s data-driven, which is why it’s a market-favorite. This framework is also often used when industry reports and broader market data aren’t available, i.e, the market (third-party) data is limited.
So, you convert existing zero-party and first-party data to analyze your current market size and potential.
You leverage existing audience data, such as offering pricing and usage, to calculate the TAM.
It helps you put the customers first. Rather than relying on a third party’s data analysis, you conduct your own market research in the arena where you know you can play best.
Count how many companies actually fit into your ICP.
Outline the average contract value (how much they’d pay you annually) for each account.
Multiply to calculate your total potential revenue, or TAM.
Example:
You’re targeting businesses in Singapore with overall 200-500 employees for your HR and payroll tech.
Number of companies in that range? – 10,000
Average contract value? – $3000
Your TAM? – 10,000*$3000 = $30 million.
All the calculations are based on actual customer and company data, and not assumptions.
With the bottom-up approach, you can negate the one dilemma that existed above- of pre-packed segments. And you can now explain to potential investors why you chose specific segments over others.
B. The Value Theory
Value theory asks, “What’s this worth?” i.e., highlights the economic value of your venture.
And mostly significant for startups with disruptive offerings.
You aren’t looking at existing markets. You are demonstrating your own worth to create a new segment driven by the value you offer. This means you’re solving an expansive market pain point that doesn’t have an established market size yet.
Value theory for TAM kickstarts investor discussions and offers market validation- what is the potential of your offering?
The process?
Outline a particular business challenge.
Measure how much this pain point can cost companies.
Quantify your offering’s value creation, i.e., how much of that cost can you eliminate.
Calculate the value of market share you can capture = total market value.
Example:
Data entry errors cost companies $750 per employee per year.
Your startup has developed a solution that can reduce the problem by 50%.
Target => 5000 companies, each with 200 employees.
What’s it costing a company? => $750*200 = $150k.
Total cost of the problem for the market ⇒ $750*5000*200 = $750 million.
How can your solution eradicate? ⇒ $750 million*0.50 = $375 million (Your TAM).
The Bottom Line- Adopt a Multi-Structural TAM Analysis Framework.
By changing the approach to TAM analysis, you aren’t eliminating the entire process but evolving it.
You must pivot to a TAM analysis framework that offers a more precise and accurate assessment of your company’s market potential. By solely sticking to the traditional method, you are limiting your understanding of your own company and its impact.
It can offer you high estimates and paint a sunny picture. But at the end of the day, it’s clouded by assumptions and doesn’t align with real-world market complexities.
A sustainable business model demands a change from this inflated TAM calculation.
Instead, opt for a multi-structural framework- one that considers diverse scenarios. And spotlights both the upsides and downsides of your market opportunities.
Alternative approaches help draw a realistic picture of a company’s market potential than depending on hearsay. Investors and founders alike can gauge its potential impact and ROI.
And highlight where you may stand in the market-
Will the market drive you, or will your offering end up driving the market?
