The astounding tripling of semiconductor giant NVIDIA in 2023 on the back of demand for the ChatGPT language model has investors very excited for stocks exposed to artificial intelligence (AI) – not just as users of the technology, but actively involved in its development.
In May, NVIDIA announced that AI-based demand for its computer chips would exceed its original sales forecast by more than 50%, as that was created by OpenAI in 2022. ChatGPT is just the start of demand for AI-related products, as technology enables a new economic revolution similar to the agricultural and industrial ones.
The AI gold rush is on. There are a handful of AI stocks on the ASX. Here is a look at five of them; and two ETFs that for most investors, might be better ways to ‘play’ the AI theme.
- Appen (APX, $2.39)
Market capitalisation: $336 million
12-month total return: –56.2%
Three-year total return: –58.1 a year
Estimated FY24 dividend yield: no dividend expected
Analysts’ consensus valuation: $2.42 (Stock Doctor/Refinitiv, 12 analysts), $1.988 (FN Arena, four analysts)
Appen is a third-party dataset provider for AI and machine learning algorithms. Appen uses crowd-sourcing methods to generate the data that goes into the machine learning models of many of the biggest tech companies in the world, used in everything from search engines to voice assistants and image recognition technology.
Appen provides data ‘labelling,’ which is the process of identifying raw data (images, text files, speech files, videos, alpha-numeric data etc.) and adding one or more meaningful and informative labels to provide context, so that a machine learning model can learn from it.
The Australian firm’s edge in this market is that it has a crowd of more than one million contributors, spanning more than 170 countries, and speaking more than 235 languages and dialects, busily evaluating, contextualising and annotating data: this diversity offers Appen’s customers advanced and curated data labelling capable of incorporating highly specific cultural, language and social nuances.
Appen works with some of the biggest users of AI in the world: names such as Google, Amazon, Microsoft, LinkedIn, Salesforce, Adobe, Oracle, Boeing and Airbus. Eight out of the ten largest global “big tech” companies by market capitalisation are clients, and its top five clients have been with it for an average of more than nine years.
The unique nature of its crowd asset, and its stature in the tech world, made Appen a market tech darling on the ASX for a long time. Listed in January 2015 at 50 cents a share, the stock became an ‘eighty-bagger,’ peaking at $40 in August 2020. Along the way, Appen became one of Australia’s so-called ‘WAAAX’ tech stocks, along with Xero, WiseTech, Afterpay (now Block) and Altium – the ASX’s answer to the FAANG grouping (Facebook, Amazon, Apple, Netflix and Google, which now trades under its parent, Alphabet).
But Appen now trades at $2.39.
Ultimately, Appen’s fortunes ride on digital advertising demand, and the spending of its largest customers to capture that demand. From 2021 onward, that demand was hammered by a combination of economic downturn, regulatory crackdown, a general sell-off in tech stocks as interest rates rose, the rise of competitive Indian outsourcing and privacy changes that Apple made to the iPhone operating system, which made it harder to track customers and target mobile ads to users.
A succession of six earnings downgrades dating from late 2020 – and a takeover approach in May 2022 by Canadian company Telus International that was pulled very quickly, when the Canadians apparently balked at another downgrade – pummelled the share price, which sank as low as $2.15 in May, before Appen announced a $60 million capital raising, at $1.85 a share, a 19.6 per cent discount, a raising that increases the number of shares on issue by 26 per cent.
The company said it would use the funds to strengthen its balance sheet and diversify its revenue to return the business to profitability, while it also looked to cut its operating costs.
The bear case is stated by broker Morgan Stanley: “that APX’s human-based model is less economical than competitors’ automated solutions.”
The company says the opposite: that the new wave of generative AI still needs humans to make it truly useful. Appen has launched generative AI-related services, such as providing human feedback for large language models, for enterprise customers. For example, Appen’s new assurance products ensure customers’ AI models are accurately measured and monitored to avoid risks such as bias, toxicity and “hallucination” – the industry’s term for when the technology sprouts complete nonsense.
Appen chief executive Armughan Ahmad has said that generative AI is an opportunity for Appen to diversify its client base, at present heavily concentrated among Google, Apple, Microsoft, Meta and Amazon. The company has announced a vision to expand beyond data for AI and build industry vertical-focused AI solutions, which it says can expand its total addressable market (TAM) 20-fold. Appen says it can win because: it has “a front-row seat on latest AI Innovation” from leading tech companies; its “deep industry vertical expertise”; its partnerships with AI ecosystem leaders; and its “execution-focused leadership.”
Ahmad has also told the market that the executive team’s remuneration is closely aligned with improving the share price, noting that they need Appen’s share almost to triple to receive almost half of their stock.
Analysts appear to be betting against the company. They expect it to make a loss in FY23 (calendar year) and FY24.
- NEXTDC (NXT, $12.53)
Market capitalisation: $6.5 billion
12-month total return: 19.2%
Three-year total return: 8.7% a year
Estimated FY24 dividend yield: no dividend expected
Analysts’ consensus valuation: $14.01 (Stock Doctor/Refinitiv, 16 analysts), $13.765 (FN Arena, six analysts)
NEXTDC is in a great position as AI demand explodes – it operates the data centres that handle the data that is created, and offers data-centre solutions such as colocation, cloud, and network services to customers in various industries such as technology, government, and financial services. The sheer scale of data capacity that AI requires will drive more demand for data centres: and NEXTDC operates, or is in the process of building, 15 data centres across Brisbane, the Sunshine Coast, Canberra, Sydney, Melbourne, Darwin and Perth, with capacity ranging from 6 megawatts to 300 megawatts.
In May, NEXTDC is raised $618 million to build two new data centres, adding sites in Malaysia and New Zealand, in the company’s first significant investment outside of Australia. And it is evaluating potential sites in Tokyo and Singapore, as it eyes further expansion opportunities outside Australia.
This makes NXT a “pure play” to get exposure to rising demand for data centres. Think of it as a kind of “selling shovels to the miners” kind of exposure to the AI theme: NXT is not looking for the gold itself, but it operates the safe warehouses that the AI miners need to store their gold.
NEXTDC has announced a string of significant contract wins in April, increasing its contracted utilisation by 43% (to 120MW) since 31 December 2022, and the stock has ridden nicely on the back of the NVIDIA enthusiasm. Until recently, NXT was one of the 15 most heavily-shorted stocks on the ASX, but the hedge-fund short-sellers have had to bite the bullet as the stock has risen NXT is up 41% in 2023 – and is no longer in the Top 15 short positions.
Having fought off the short-sellers, analysts like the look of the NEXTDC price at these levels; there appears to be reasonable potential upside. The company does not pay a dividend.
- Weebit Nano (WBT, $5.06)
Market capitalisation: $956 million
12-month total return: 112.6%
Three-year total return: 167.6% a year
Estimated FY24 dividend yield: no dividend expected
Analysts’ consensus valuation: n/a
Weebit Nano, an Israel-based semiconductor components company, has had an interesting 2023. The company is trying to produce a new kind of semiconductor memory chips – but it has not yet produced any revenue. There was nil revenue in FY22, now was there in the half-year to December 2022; although Weebit says the company had “achieved several key technical and commercial milestones” during the half, and it expected to bring in its first sales revenue in 2023.
Weebit’s product is resistive random-access memory (ReRAM). It is non-volatile memory (NVM) technology, which is semiconductor memory that stores information even when disconnected from electrical sources, like a USB card or hard drive. It’s an advance on ‘volatile’ semiconductor memory – such as dynamic RAM (DRAM) and static RAM (SRAM) – in which data is only retained so long as a device is connected to a power source.
In particular, it is an advance on “flash” semiconductor memory, which has been a widely used NVM technology, but as the semiconductor industry follows Moore’s law to ever-smaller chips, the “flash” chips have got about as small as they can get, and they’re no longer fast enough.
Weebit’s ReRAM has ultra-low power consumption, excellent retention even at high temperatures, fast access time, and has high tolerance to harsh environmental conditions, including high temperatures, radiation and electro-magnetic interference (EMI). This makes the ReRAM technology ideal for applications such as Internet of Things (IoT) devices, smartphones, robotics, medical equipment, autonomous vehicles, 5G communications. The company’s technology will also be suitable for future ‘embedded’ applications, such as “edge AI,” where data is processed and acted upon at “the edge” – the location where the data is created – rather than sent to the “cloud” in a data centre and retrieved for processing.
As 5G and AI change our society, NVM is a critical component that is enabling this paradigm shift. According to tech research firm Allied Market Research, the global NVM market size was valued at US$41 billion in 2022 but is projected to reach $96.1 billion by 2032.
Weebit’s technology realisation partner, the Nasdaq-listed SkyWater Technology, has been putting the products through all sorts of manufacturing and performance tests (warning: there is eye-watering jargon in WBT’s ASX announcements), but the gist is that it’s ready to go to market, and to commercial orders. This is what Weebit shareholders are backing, and they’ve determinedly pushed the share price higher this year, through short-selling attacks from hedge funds and a big equity raising. WBT is up 46% in 2023 alone. It is hard to find analyst coverage on this stock, but it may have been pushed too far.
- BrainChip (BRN, 37.5 cents)
Market capitalisation: $672 million
12-month total return: –53.1%
Three-year total return: 66.7% a year
Estimated FY24 dividend yield: no dividend expected
Analysts’ consensus valuation: n/a
Australian company BrainChip is another whose ASX announcements are very difficult to read for the lay investor. The company bills itself as the worldwide leader in “edge AI on-chip processing and learning” – as mentioned above, “edge” means “the location where the data is created.”
BrainChip is working to commercialise its Akida neuromorphic processor. “Neuromorphic” means that the processor uses neuromorphic principles to mimic the human brain: it analyses only essential sensor inputs at the point of acquisition, and BrainChip says Akida continues to learn from experience and evolve, autonomously, like the human brain. This means that Akida can process data with “unparalleled efficiency, precision, and economy of energy,” according to the company
Another term for it is a “spiking neural network.” But in simple terms, when integrated into computer chips, Akida works as a fully digital processor that gathers AI through event-driven processing to learn in real-time, in the same way as the human brain does, and can deliver AI reasoning and conclusions from sensor-captured data, at the “edge.” Being at the edge, Akida can “learn” very quickly, at the chip, and independent of the cloud – this means it does not need a continuous internet connection, and thus requires far less power.
Edge computing takes what are “dumb” devices with sensors that send data to the cloud (data centres) and turns them into “intelligent” devices capturing and processing data at the point of capture, that can be acted on immediately and efficiently. BrainChip says “devices are the new computer” – and the “artificial intelligence of things (AIoT) market – the combination of AI technologies and the internet of things (IoT) infrastructure will be beyond massive.
Specifically for the edge market where operates, BrainChip cites Forbes Business Insights as stating that Edge AI technologies will accelerate the AIoT market, which is conservatively estimated at more than US$1.2 trillion ($1.8 trillion) – that market includes a large addressable market for AI-enabled edge and endpoint silicon solutions, like BrainChip’s Akida.
But there are several flies in the ointment.
First, global AI leader NVIDIA also has products that cover edge AI – the same market that BrainChip is targeting.
Second, BrainChip has not been able to get a commercial product to market.
In May, chairman Antonio Viana told the annual general meeting: “Let me be clear, nobody at BrainChip is happy or content with our current position. We haven’t hit any significant stride yet with respect to revenue. No-one is satisfied, and no-one should be. However, there is plenty to be positive and optimistic about. BrainChip has done more in the past 12 months to strengthen its development, market position and talent than it has since its inception.”
The company told shareholders that while its engineering team was working very hard to complete the second-generation Akida product, it was already planning the third generation.
It is a fascinating story – again, it is very difficult for a lay person to understand the company’s ASX announcements and communications – but there is very little analyst coverage that is easy to find. Stock Doctor/Refinitiv does not have any analyst target prices on the stock; nor does FN Arena. Investors really have to do their own assessment on BrainChip – and whether its technology can capture meaningful share of a potentially colossal global market.
- Archer Materials (AXE, 61 cents)
Market capitalisation: $155 million
12-month total return: 10.9%
Three-year total return: –0.6% a year
Estimated FY24 dividend yield: no dividend expected
Analysts’ consensus valuation: n/a
Archer Materials is developing the next generation of semi-conductors, using innovative technology that uses carbon-based “qubits” rather than traditional silicon-based semiconductors for quantum computing. A qubit is the fundamental unit of information in quantum computing; the physical properties of qubits enable quantum computers to perform some computations much more efficiently than the computers available today. Some of these specific tasks may also include computations related to artificial intelligence and machine learning.
With its ability to powerfully process vast amounts of data and then quickly and accurately test and model complex systems, quantum computing is likely to enable future generations of artificial intelligence.
Archer is essentially a nanotechnology company, that has synthesised single-atom graphene (a carbon nano-material) and built a single-atom transistor. Qubits can be implemented using various quantum systems, however there are only a few available today, and include superconducting circuits that need to be cooled to ultra -low temperatures to operate: a major benefit of Archer’s approach is that it removes the need for the extremely low temperatures required by silicon-based quantum computing.
Qubits are very sensitive to environmental disturbances, such as heat and magnetic fields – what is referred to as ‘decoherence’, that ultimately leads to errors in quantum computations. Protecting against decoherence is one of the major challenges in developing quantum processing chips. Archer has made major breakthroughs in this regard: its qubit material has shown the potential to facilitate quantum computing at room temperature while also integrating easily with modern electronic circuits, dramatically streamlining the complex infrastructure that has limited the accessibility to quantum technology.
While Archer’s technology has a huge range of potential applications in quantum computing, it is initially focusing on “lab-on-a-chip” biosensors for integrated bioelectronics; it describes this as effectively miniaturising medical labs onto a chip.
Continuing the refrain of this article, Archer Materials is a very difficult stock to self-research – and you don’t get much help from analysts.
- BetaShares Global Robotics and Artificial Intelligence ETF (RBTZ, $13.18)
Market capitalisation: $199 million
12-month total return: 47.4%
Three-year total return: 8.4% a year
Historical FY22 dividend yield: 0.9%
Analysts’ consensus valuation: n/a
- Global X Global Robotics and Automation ETF (ROBO, $76.85).
Market capitalisation: $257 million
12-month total return: 33.7%
Three-year total return: 12.1% a year
Historical FY22 dividend yield: 8.1%
Analysts’ consensus valuation: n/a
The most efficient, and most diversified way for Australian investors to ‘play’ AI comes in the form of two specialised ETFs, BetaShares’ Global Robotics and Artificial Intelligence ETF and Global X’s Global Robotics and Automation ETF. This pair offer investors access to portfolios of the world’s leading robotics and AI companies, in one stock. (NVIDIA is the top holding of RBTZ, accounting for 12.2% of the portfolio, while the stock is only the 26th largest holding in ROBO, at 1.4%.)
As with any ETF, you get the winners in the space, and the laggards – what you get is broad exposure, in this case, to the global AI industry. But the diversification and ease of access of the ETF takes single-stock risk out of the equation.
Important: This content has been prepared without taking account of the objectives, financial situation or needs of any particular individual. It does not constitute formal advice. Consider the appropriateness of the information in regards to your circumstances.
